Multimedia Tools and Applications (2023) 82:8983–9027

https://doi.org/10.1007/s11042-022-12167-w
1209: RECENT ADVANCES ON SOCIAL MEDIA ANALYTICS AND MULTIMEDIA
SYSTEMS: ISSUES AND CHALLENGES

Tourism recommendation system: a survey and future

research directions

Joy Lal Sarkar1 · Abhishek Majumder1 · Chhabi Rani Panigrahi2 · Sudipta Roy3 ·
Bibudhendu Pati2
Received: 6 November 2020 / Revised: 11 November 2021 / Accepted: 7 January 2022 /
Published online: 21 April 2022
© The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2022

Abstract
A Recommendation System (RS) is an intelligent computer based system which provide
valuable suggestions to the user and are used in several domains. Social media platforms
are the most common internet applications due to the large number of users. The numerous
posts, likes, etc. have accrued on social media platforms and can be used in variety of
recommendation systems. In this work, the primary focus is the tourism domain, where RS
serves as a valuable tool for the tourist to plan his trip. Traditional RS systems only cater
to the needs of the tourist by examining few factors. However, there are a large range of
factors such as environment factors , actual geo-coordinates, trip destination, preferences of
the user etc. that need to be taken into account in order to make a foolproof recommendation
to the tourists. Tourism Recommendation Systems (TRS) provide suggestions to the tourists
to identify the most suited transport (flight, train, etc.), accommodations, museums, special
interest places and other items which are required for the trip. Several techniques are used
and a thorough study of various techniques of traditional RS and TRS techniques have
been done which are specially designed for tourism domain. Various Artificial Intelligence
(AI) techniques have been highlighted which are used to solve the tourist recommendation
problem. Also, future research direction has been suggested which would improvise the
Quality of Service (QoS) of the RS in tourism industry.

Keywords Tour recommendation · Social media · Tourism · POIs · Itineraries · RS · AI

1 Introduction

Recommendation System (RS) is an intelligent computer based system which serves as a

guide and suggests, as per the preferences of the person [98, 163]. It uses state of the art

 Abhishek Majumder
abhi2012@gmail.com

Extended author information available on the last page of the article.

8984 Multimedia Tools and Applications (2023) 82:8983–9027

techniques like Big Data, Artificial Intelligence (AI) etc., and benefits both the consumer
and the merchant [93]-[114]. RS is becoming very popular as it serves as a guide, for the
activity that a person or a group plans to perform in the best possible manner, given the
constraints imposed by the user(s). In the tourism industry, a tourist has numerous options
and can easily be lost. However, the RS comes to the rescue as a guide, by exactly sug-
gesting about the various amenities that he/she needs to take in order to make the trip a
memorable one [119]. RS have been present in the market, since the 1990’s [87], and are
constantly being upgraded by adding new techniques of machine learning and state of the
art technologies like Smart Mobile Device (SMD), Big Data, etc. RSs are computer based
intelligent techniques which solve problems of information over-loading, and are not search
engines. RS is beneficial both to the consumers and the merchants and enhances user expe-
rience by assisting users in finding products information as per the taste and needs and also
benefit the business by increasing the sales [97]. The key task involved in RS is to quan-
tify how much a user likes an item and then predict the list of items that a user is most
likely to buy. The approaches used in RS are either Content based filtering (wherein the
system recommends the items that are similar to the items browsed by the user), Collab-
orative filtering and Hybrid filtering which is a combination of both the Content and the
Collaborative based filtering. Various complex techniques such as heat conduction and mass
diffusion are also being used in RS which have been introduced through advanced research
done by physicists and complex system scientists [162]. An abstract structure of the social
network involves different relationships between people, such as friendships, mutual inter-
ests and shared knowledge. Social networking services can be used to explore the actual
social links between different users. It enables the users to exchange ideas, activities, likes,
dislikes and news. These data can be used by different RSs for future recommendation of
the items. Tourism Recommendation System (TRS) is a specific RS for the tourism indus-
try and provides valuable suggestions and guidance to the tourists in identifying amenities
(like transport, hotel, attractions and special interest places) as per the taste, interest, likes
and budget to make the trip a memorable one. The various AI techniques used in TRS
are Knowledge Representation (KR), Optimization techniques, Clustering, Multi-agent sys-
tems, etc [12]. The popular types of TRS are Web based and Mobile based, however, there
are other types which need to be downloaded on desktops / laptops or on mobile devices and
are not so popular, since tourists like to use simple systems without getting into much tech-
nical aspects. Web based TRS is commonly used in the e-Tourism industry as it presents a
user-friendly interface providing images, maps, videos and good amount of data which the
tourist can drill down as per interest. Mobile recommender systems (MRS) is gaining pop-
ularity as there are 7.2 billion mobile devices across the globe today, which is far more than
the human population of around 7.19 billion [97]. A good example of MRS is a TRS which
is installed on a SMD, and guides the tourist about the various necessities during the travel.
Also, the cab drivers can easily find places and routes having less traffic so that the tourists
need less time to travel to the places of attraction.
In this work, the authors have summarized the research on tourism recommendation sys-
tem which will provide a strong foundation to the future or young researchers in the area
of tourism domain. In true sense, the work will be helpful for designing a system which
provides practical guidance to the tourist in order to make his/her trip a memorable one.
The importance of tourism in today’s world is immense as it is a big source of revenue and
employment generation for a country. Hence, the economy of a country is elevated by the
inflow of greater number of visiting tourist.
Multimedia Tools and Applications (2023) 82:8983–9027 8985

1.1 Motivation

The primary motivation for conducting this study is to explore prominent TRS developed
so far and identify/highlight future research directions that help making TRS more robust,
personalized and propose cost-effective suggestions to tourists. Tourism is a very popular
industry with many advantages. It helps the country to earn revenue and also brings devel-
opment in the particular area by promoting world class modern facilities to be present in
that area [12]. Presently, the number of tourists has grown many fold, as people are becom-
ing more global and exploring the world. Globalization of businesses has also boosted the
tourism industry, as well as friends and relatives who are now spread in almost all corners of
the world which has resulted in lots of travel by several persons around the year [12]. This
phenomenal boost to the tourism industry demands a TRS to fulfill the specific needs of the
tourist in a cost efficient manner. Tourism data can be taken from various Internet resources
and travel agencies. Most of the travel agencies and Internet resources recommend popu-
lar attractions only. The recommendations may not fulfill the user expectation/interest. To
overcome these problems, the current research community proposed numerous alternative
approaches and techniques. Figure 1 shows the overview of the survey carried out in this
work. The figure also shows that the work is mostly focused on the TRS. The work dis-
cusses the challenges, factors, features. The primary focus is on RS techniques that are used
to handle tourist recommendation problems, as well as other strategies that are specifically
built for the tourism domain. A number of AI techniques have also been highlighted that are
being explored to solve the tourist recommendation challenge.

1.2 Contribution

The contribution of this work are as follows (see Fig. 2):

– A comprehensive overview of RS and their different categories are mentioned in this
work and also, explore the usage of RS in TRS.
– The different functionalities of TRS have also been discussed here.
– The necessity of TRS is also presented in this study.
– Highlighted the use of a variety of social media sites in the field of tourism.

Fig. 1 Overview of the Survey

8986 Multimedia Tools and Applications (2023) 82:8983–9027

Fig. 2 Review methodology of TRS

– highlighted the tourism relevant problems which could be resolved by conventional RS.
A comprehensive overview of the existing works which are specially designed for the
TRS domain and also some of the AI techniques which are used in TRS are mentioned
in this work.
– Categorize the discussed works based on the previous step.
– Towards the end, a few research directions have also been high-lighted with respect to
the existing TRS techniques.

1.3 Article structure

The remainder of the paper is structured as follows, Section 2 discusses the comparison
of related survey articles. Section 3 includes overview of RS and TRS. The functionali-
ties available in TRS is presented in Section 4. Section 5 includes the need for tourism
recommendation systems. TRS using social media analytics is presented in Section 6.
Section 7 presents the techniques available in TRS. Future research directions are discussed
in Section 8. Finally, Section 9 concludes the article.

2 Comparison of related survey articles

The work is different from the existing surveys [12, 87, 156] in many aspects (see Table 1).
(i) The earlier works discussed about the various AI techniques, which were used to solve
the TRS problem. (ii) Existing survey works are mostly concerned with the tourism indus-
try. Some of the recent surveys on TRS discuss about the personalized and group-based
recommendation systems in the domain of Tourism. In this work, the primary focus is on the
RS techniques which are used to solve the tourist recommendation problems and also some
of the techniques which are specially designed for Tourism domain. Various AI techniques
have also been highlighted which are used to solve the tourist recommendation problem.
The main objective is to find the answer to the following question - Are the RS techniques
good enough to provide recommendations to the tourist, or some other advanced tech-
niques are required which may be used specially for the Tourism domain?. These answers
can be obtained only after surveying the existing works in this area. The reviewed articles
were gathered from many sources, including Web of Science, Google Scholar databases,
ScienceDirect, and others.
The articles published in the field of tourism are reviewed before carrying out the liter-
ature review. Number of papers that offer different views on TRS were the outcome of this
initiative (see Fig. 3) ([122]).
Multimedia Tools and Applications (2023) 82:8983–9027 8987

Table 1 Summery of the existing related works

Author Contributions

[12] Several features of these systems are examined, including their interface functionality,
recommendation processes and AI approaches and techniques used.
[87] This study examines current recommender system application advances and categorizes
their applications.
[156] Offer a brief assessment of the task of POI recommendation in Location Based Social
Networks (LBSNs) outline potential research directions for POI recommendation

3 Overview of RS and TRS

This section covers the basic idea about RS and TRS. Recommendation techniques are
mainly classified into Content Based Recommendation System (CBRS), Collaborative
Filtering Based Recommendation System (CFRS), Knowledge Based Recommendation
System (KBRS), Hybrid Recommendation System (HRS). However, recently there are
other types of techniques used for recommendation, such as Computational Intelligence
based Recommendation System (CIRS), Social Network based Recommendation System
(SNRS), Context Awareness based Recommendation System (CARS), and Group Based
Recommendation System (GBRS).
1) CBRS takes into account the content of products and helps the users by suggesting the
appropriate item(s) depending upon the interest profile (Up ) of the particular user and
the brief detail of the item. In CBRS, each attributes of an item needs to be compared
and consequently, the best item is recommended to the user which has the high degree
of similarity. The CBRS fails to provide the best suitable recommendations in case

Fig. 3 Distribution of tourism recommendation system articles by years

8988 Multimedia Tools and Applications (2023) 82:8983–9027

sufficient information is not available about the user’s likes and dislikes. Up maintains
information about the user interests and likes [110] which are detailed as follows.
(a) There may be a large number of descriptions available about user’s interest items
but there needs to be a common representation function (f ). For each item that
the RS predicts there exists a likelihood that the use may be interested for the
particular item.
(b) A RS will contain the history of users’ intersections which helps to recommend
the particular item to the user with greater accuracy. For example Up will be made
when the user purchases any item (s) or the rating has been given for that particular
item by the user. Also the history of the different types of queries will be stored.
The parts of Up will be shown as a CBRS whereas Amazon has a feature named favorite
which represents the details of each item categories which is preferred by an user.
2) CFRS is a useful technique used for deriving the recommendations which are suitable
to a target user, and is based on the opinion of similar users having the same inter-
ests [105]. It is mainly of two types, memory and model based [131]. The model based
CFRS has two different phases. In the initial phase, matrix containing ratings given by
the user are used for generation of the model which represents the original matrix itself.
subsequently, in the next step the model which is generated serves as an input to cal-
culate the rating of predictions generated for a particular user. In the second phase of
CFRS, the total matrix is used to calculate the prediction and can be split further into
two approaches, which are user based and item based. In the user based approach, the
recommendation is made on the basis of the preferences of another user who has sim-
ilar tastes and likes. In case of an item based approach, the recommendation is made
based on the user’s past experience and preference of the particular item.
3) The user preferences are taken into consideration in KBRS and recommendation is
provided based on the prior knowledge about the items, such that it suits the user’s
requirements. The knowledge is present in encoded form and serves as input data to the
system. There are three varieties of knowledge which are fed into the system, namely,
Catalog, Functional and User knowledge [17]. Knowledge Base (KB) is one of the key
component of KBRS. Depending upon the quantitative and qualitative content of the
KB, the KB is paired along with the recommendation strategy by involving similarity
measures and matching techniques [17] and a Up is created which is composed of the
preferences of user, tastes, user interests, requirements, etc. The content of Up depends
on the methodologies as well as the strategy of recommendations.
4) It is also possible to build a RS wherein two or more different techniques can be inte-
grated. In case of NewsDude, it normally uses the K-nearest neighbors (KNN) and
Naive-Bayes classifiers for recommendation [10]. Burke et al. [17] describes seven
different types of HRS techniques, namely Weighted [100], Mixed [130], Feature
combination, Switching [10], Seature augmentation [106, 144] Meta-level, and Cas-
cade [17, 109]. In the weighted technique, the votes or scores of different techniques
used for recommendation are grouped in order to make a final recommendation to the
user. In case of Switching technique, it is possible that different technique for recom-
mendation are applied based on the present circumstances. In the Mixed technique,
the recommendations are generated from different recommender’s at the same time.
In the Feature combination technique, the different features derived from various data
sources involved in HRS, are aggregated to form a single algorithm. In case of Cascade
technique, the output from a recommendation engine is refined by the subsequent RS,
Multimedia Tools and Applications (2023) 82:8983–9027 8989

thereby providing better recommendation to the user. In the technique, based on Fea-
ture augmentation, the output of an algorithm serves as the input for the next algorithm.
In the Metalevel technique, a HRS is called Meta-level if the model is learnt using a
particular recommendation algorithm and the output of the first algorithm is an input
to the next one.
5) Dey et al. [29] has stated that “context” as the information required to define an entity.
Entity can be an object (like person, place etc.) and is used for interactions with an user
or an application
Based on a system which has full information, a RS can be categorized into three
categories namely Fully observable, Partially observable and Unobservable.. In case of
a RS which is of the type of observable fully, the value and the structure are explicitly
known, A Partially observable RS has few contextual factors that are explicitly known.
6) CFRS offers more accurate recommendation to the users, however this technique
suffers from various drawback like cold start, scalability and sparsity,which leads
to inaccurate recommendations. In order to overcome these issues, it is suggested
that CFRS be integrated with other computationally intelligence techniques such as
Bayesian classifier, k-means clustering, Artificial Neural Networks (ANN), Fuzzy set
techniques etc. and is called as CIRS. Bayesian classifier is used for solving the prob-
lem of classification which is based on the conditional probability definition as well
as Bayes theorem and is useful for model-based RS. ANN consists of an assembly of
nodes which are interconnected and the weighted links which are used for developing
the model-based RS [4]. Hsu et al. [51], devised a Television RS using the ANN tech-
nique which uses the back-propagation NN (Neural Network) technique for training a
NN which has three hidden layers. Clustering techniques are useful for grouping the
similar items which minimizes the computational cost, like in the case of K-Nearest
Neighbors (kNN) algorithm. G-r et al. [37], came up wih a RS approach wherein
the clusters were used for dealing successfully with the unrated data for each of the
users. GA is useful for parameter optimization problems [59]. GA is mainly used for
clustering [59], and hybrid user models [1]. A GA technique can be used for obtain-
ing the function of the optimum similarity and it is clear that this function provides
higher quality as well as faster results with respect to the traditional metrics. The fuzzy
set techniques have been used when the necessary information about any situation or
domain is incomplete.
7) The number of users who are using the social network is increasing day-by-day. Based
on the analysis of social network, an intelligent RS can be made which can provide
more interesting recommendations to the user. SNRS can provide the opportunity to
users to interact with other users using the social network techniques such as social
tags, comments, etc. SNRS fully utilizes the information available in social networks
which boosts up the performance of traditional RS. The logic behind this technique is
to utilize the simplified homogeneous social networks as a Bayesian network.
8) The large majority of RS are designed to make recommendations for individual
users [166]. The items suggested to a particular user may not be intended for the single
user, but may be recommended for a set of users. GBRS helps by providing useful sug-
gestions to a set of users when members of a group are unable to interact face to face
or do not have clarity about their preferences [41, 165]. GBRS is used in various areas
like Entertainment and various complex scenarios.
The advantages and disadvantages of various Recommendation techniques are shown in
Table 2.
8990 Multimedia Tools and Applications (2023) 82:8983–9027

Table 2 Summary of Recommendation Techniques

Techniques Advantages Disadvantages

CBRS [110] • User independence: exploits solely the rat- • High scoring: it suggests only the high
ings which are given by individual user to scoring items which matched against Up .
setup their profile
• Transparency: The recommended list of • Overspecialization: there exists no inher-
items that are provided by the CBRS are ent method in CBRS for finding any unex-
clearly explained based on the detailed pected scenario.
working of the algorithm thereby providing
sufficient explanation for the same.
• New item: CBRS can provide recommen- • New user: to understand the user prefer-
dation even if the user has not provided ences and make recommendations, system
any rating for the particular item. Hence the should collect enough ratings from the user.
first-rater problem does not exist. On the contrary, the system fails to pro-
vide suitable recommendations to a new user
when less amount of ratings are available.
CFRS [131] Memory based: Memory based:
• New data can easily be added • Determined by human ratings
• Implementation is easy • Cannot provide recommendation for a new
user
• No need to consider content of an item • Limited scalability in case of large datasets
• The scalling is done with respect to the • Performance gradually decreases when data
corated items is sparse
Model based: Memory based:
• Prediction performance can be • Trade-off between prediction performance
improved and scalability
• Addresses the issues related to sparsity, • Useful information can be lost due to the
scalability, etc. application of dimensionality reduction
techniques
• Provides an intuitive rationale for better • Expensive to model
recommendation
HRS [10] • Overcomes the drawback of CBRS, CFRS • Due to the integration of two or more RS the
and other RS by integrating two or more RS complexity and expense may be increased
• Prediction performance can be improved • External information is required which may
not be available
• Overcomes sparsity and gray sheep
problem
CARS [29] • When the user interacts with a particular • Most of the techniques used in CARS center
context, this technique helps to resolve the around the context, whereas the other
problem of “situated actions” alternative methods are under-explored
• Prefiltering, postfiltering, and contextual
modeling paradigms are partially explored
KBRS [17] • Can map the user requirements to • The capability to learn on its own does
products not exist
• KBRS include non-product features • There is a need for knowledge engineering
as well to be incorporated in KBRS
• Ramp-up is not required
Multimedia Tools and Applications (2023) 82:8983–9027 8991

Table 2 (continued)

Techniques Advantages Disadvantages

SNRS [159] • Improves the accuracy of recom- • user who have less interest in net-
mendations which are dependent working with others do not contribute
on the social interaction pattern of significantly in such type of RS
users
• There is no need to find out the • Recommendations provided to the
similar users users who are not active in social net-
working may not be accurate
GBRS [165] • Solves the cold-start problem, • There exists very large degree of com-
wherein a suitable recommendation plexity to provide recommendation to
can not be provided due to lack of the group of users
information
• Dealing with uncertainty
CIRS [51] • Overcomes the issues related to • Most of the time computational intel-
CFRS like cold start, scalability and ligence techniques is very complex to
sparsity design and can only be handled by the
expert as black box approaches
• Over-filtering wherein the perfor-
mance is good for training dataset but
poor performance for the test dataset
when a model become very complex.

3.1 Tourism recommendation systems

TRS provides suggestions to the tourists to identify the most suited transport (flight, train,
etc.), hotel, restaurant, map guidance and other items which are required for the trip. The
recommendation for the trip and planning of itinerary are difficult tasks for the tourist.
Point-of-Interest (POI) is an important parameter to design a TRS. POI refers to a definite
location point which the user finds interesting in some aspects [33]. There are various factors
namely tourist interest on a particular POI, popularity of POIs, time or distance travel costs,
transport/traffic conditions etc. which play an important role during recommendation of
POI.
A personalized POI consists of two basic elements such as the information related to the
particular POI and the user [61, 155].
1) Attributes related to POI: The information can be split into subjective and objective
attributes. In case of subjective attributes, the information is collected using the visitor’s
opinion for a particular POI, whereas, the objective attributes include the information
about the tourist such as location of POI [11].
2) User attributes: The user attributes affect the POI selection which consist of user’s
personality, behaviors, social factors, etc. A particular user may tend to select different
POI with respect to another similar user, even though the same tourist information is
presented to two different users.
The attributes can be further split into static and dynamic types. The static attributes need
input data only during initial registration. In the dynamic attributes the user’s input is con-
sidered at all times. The recommendation is continuously changing because of the changing
condition of the user during a tour [33].
8992 Multimedia Tools and Applications (2023) 82:8983–9027

The POI recommender systems are different from the traditional RS in some aspects
which are detailed as follows [156].
1) Geographical Influence (GI): This particular characteristics of GI distinguishes the tra-
ditional RS and POI recommender system. In general, a user may be interested to
visit nearby places which indicates that a user may visit several POI’s surrounding a
particular POI.
2) Frequency data and sparsity: In traditional RS, a user generally expresses his/her inter-
est using ratings, where a higher rank indicates higher satisfaction level and may form
User-item Rating Matrix (URM). Unlike to traditional RS in POI recommender sys-
tem, a user’s preference is made based on his/her frequency of visit to various locations
which is termed as User Location Check-in Frequency Matrix (UCFM). The sparsity
of ULCFM is larger than URM.
3) Social Influence (SI): To increase the recommendation quality, traditional RS combines
the social relationship between friends with ratings, but in [48], the authors have shown
that, 96% of users share less than 10% of commonly visited places of interest.
4) Temporal influence: In case of POI recommendation, temporal influence plays a vital
role because during the check-in activity, there might be physical constraints as well.
Generally, in case of POI, there are mainly three aspects of temporal influence, namely,
consecutiveness, non-uniformness, and periodicity [38]. The behavior of user’s check-
in exhibit a periodic pattern because a user may check-in restaurants at noon but also
have fun at nightclub. On the other hand, the non-uniform aspect of temporal influence
depicts a user’s preferences of check-in which may vary depending on the specific hour
of the day, specific month of the year, etc.
5) Content Indication: A user may generate contents, including POI ratings, geo-tagged
photos, etc. [112]. It is possible that the related content might not be tagged to each
of the check-in record. The user comments may be used to improve the degree of POI
recommendation.
A proper recommendation can be made using a suitable itinerary planner which is basi-
cally a computer software program. The itinerary planner is completely different from the
exiting trip planner in three aspects [78]. Firstly, the itinerary planner has considered multi-
modal itineraries, secondly, the itinerary planner has ability to handle multiple constraints
and also the time windows for arrival as well as departure time. Finally, in order to recom-
mend the best itinerary for a particular user, the itinerary planner should adopt an interactive
programming approach. Moreover, to designing an itinerary planner, there are two chal-
lenges. Firstly, on which criteria itineraries can select by the planner. Generally, in case of
real-world situation, a single itinerary is not sufficient and to select itineraries, a planner
must consider multiple criteria and trade-offs among them. On the other hand, the second
challenge is constraints which are associated with the trip. These constraints are imposed
by travelers or exist as intrinsic attributes of a particular visit.
For modeling personalized tourist guide, there are three functions that need to be con-
sidered, namely, the profile of a tourist, route generator, and finally, customization [78]. In
addition, the tourist factors are dependent on the different characteristics such as social-
demographic, geographic, etc., and tourist’s preferences such as POI, time and money
budget, etc. [47] discussed about the Tourist Trip Design Problem (TTDP). There are
three methods available for TTDP, namely, the profitable Tour Problem (PTP) which maxi-
mizes the profit. The Prize Collecting Traveling Salesman Problem (PCTSP), minimizes the
Multimedia Tools and Applications (2023) 82:8983–9027 8993

traveling cost with profit of the total tour and is not less than the actual value. The Orienting
Problem (OP), maximizes the profit, keeping travelling cost below a certain value.
Generally, TTDP problem involves with problem of the route planning in case a tourists
plans to visit multiple POI.
When the tourist wants to visit multiple places which are his/her POI, the route covering
all places needs to be planned efficiently, such that he/she is able to cover all places in an
orderly manner keeping in view the constraints such as time required, distance between the
different POI’s, cost, etc., and also ensuring that the satisfaction is maximized and the trip is
really a memorable one at the end of the day. This aspect of the trip is termed as the TTDP
and involves intelligent algorithmic approach and methodologies in order to resolve the trip
design issues using the constraints of the tourist and the different attributes of the POI. The
new prospects in trip planning problems are the arbitrary start / end tour locations, solving
realistic TTDP formulations, fast tourist routes updates and parallel computation.
Keeping in mind, the various constraints of the tourist and the POI attributes, ordered
visits are planned by the daily tourist tours using TTDP. The main aim is to select a POI
which matches the preferences of the tourist leading to maximization of a satisfaction of
the tourist, and also taking into account the various parameters and constraints. The aim of
this work is to survey the various models, algorithm and methodologies which are related
to the tourist trip design. This work also examines the recent approaches, processing on the
models which can capture a multitude of realistic POI attributes in the best possible manner.
Different open issues and the promising aspects of future research work have also been
discussed.

3.2 Summery and insight

Tourism is a kind of activity that strongly connects tourist’s interests and preferences. In any
case, it is difficult to apply RSs to the tourism domain. A lot of components must be ranked
in order to apply CBRS (see Table 12). Generally, a tourist can’t rank a lot of components.
CFRS is a helpful method used to derive suggestions that are appropriate for the target user
and is based on the view of comparable customers with the same interest. However, CFRS
can not make a suggestion for a new tourist. It is also feasible to construct a RS in which
two or more distinct methods can be incorporated. Due to the inclusion of two or more RS,
the complexity and expenditure may be improved. On the basis of an assessment of the
social network, a smart RS can be produced, which can make more interesting suggestions
to the customer. Users who have less interest in networking with others do not make a
significant contribution to this sort of RS. Generally, a person travels with different people,
so the inclinations of a couple of tourist ought to be borne as a main priority. GBRS helps
to provide helpful suggestions to a number of tourists. It is evident from the discussion in
Section 3 that the traditional RSs are not adequate to fix the TRS issue. Therefore, there are
so many works specially intended for tourism domain.

4 Functionalities available in TRS

There are various functionalities available in TRS such as Trip Planner, Social Aspects,
Tourist Packs and Suggested Attractions [120]. These functionalities are discussed below.
– Tourist Packs: Seidel et al. [127] provides recommendation regarding the tourist desti-
nation and purchase of different services such as booking of the trip from their personal
8994 Multimedia Tools and Applications (2023) 82:8983–9027

travel agents. Lorenzi et al. [85] used travel agencies to give best packages on the basis
of user preferences.
– Suggested Attractions: TRS suggests numerous restaurants, hotels, attraction location,
etc. to the tourist when he/she has already fixed the destination point or availability
places. Ruotsalo et al. [118] proposed SMARTMUSEUM which can easily detect when
a user is present either indoor or outdoor. Gavalas et al. [46] developed a TRS based
on collaborative filtering techniques by considering the user’s current location, weather
condition and also the already visited places. The authors have installed Wireless Sensor
Networks (WSNs) on the user side for uploading user ratings. [83] analyzed the chat
messages between the tourist and the travel agent. The system automatically searches
the database and retrieves the user specific elements such as user attraction or cities
based on their preferences.
– Trip Planner: Lot of work has already been done provides list of places from which
the user selects their target places according to their needs and also to creates a route
combining several attractions. Huang et al. [55] proposed an ontological based For-
mal Concept Analysis (FCA) approach in order to integrate the heterogeneous tourist
information for online tour planning.
– Social Aspects: A good amount of work has been done based on the social functionali-
ties such as picture, comments or videos etc. and interaction with other users. Kosmides
et al. [63] developed a recommendation technique based on machine learning. Authors
have used the social networks data in order to produce best recommendation system
for the user. In [57], the authors have used the information of the social network in
order to improve the Top-k recommendations and has been improvised using the Matrix
Factorization and Nearest Neighbor algorithms [150]. In [138], the authors have pro-
posed a tourist recommendation system named as GuideMe which integrates the social
networks and the unique set of options which are given in the applications.

5 Need for tourism recommendation systems

In case of TRS, the main actor is a tourist who is exploring new places. Tourism is mainly
classified into various categories such as medical, educational, etc. all these domains consist
of a large number of information which are stored digitally. This work addresses various
technical challenges where traditional RSs are not sufficient to solve the problem of TRS.
– Dynamic Itinerary Planning: To plan an optimal itinerary is a great challenge for
tourists. Generally, a tourist may visit different places in a city, restaurants etc. The
existing RSs recommend tourists based on tourist’s interests, but their nature are static
which means, the systems ignores the real-time changes of different places.
– Mobility of Users: The prevalence of mobile technology in the world has boosted the
popularity of Mobile RS (MRS). Today, SMD will be present in every human being’s
hand, and often more than one for a single person. Life has become unimaginable with-
out a SMD for people across the globe today. The advantage of using MRS is that the
tourist can use it anywhere as SMD’s have an internet connection and access to infor-
mation is there at all times, which provides up to date suggestions to the tourist keeping
him abreast of the situation of the particular amenity or the attraction. Also, the SMD’s
have GPS which provides Geo-referenced information and guide the tourist to the right
attraction as per his interest. It is also necessary that all relevant information is squeezed
Multimedia Tools and Applications (2023) 82:8983–9027 8995

in the well-structured manner onto the screen of the mobile device (whether it be a
mobile phone, tablet, laptop, etc.) to help the tourist in fast manner.
– Evaluation Methods: Most of the techniques for RS consider tourist’s feedback for
evaluation methods using Mean Average Error (MAE) and Root Mean Squared Error
(RMSE).
– Integration: The main challenge in TRS is the seamless integration between the var-
ious sources of information such as airline, restaurants, and travel etc. It is required
to integrate this information in a condensed format. For example, when trying to
extract the information about the preferences of individual tourist, it is necessary to
get the information from various systems. The integration plays a mazore role in this
case.
– Serendipity: Serendipity is the concept of exploring a new interest which the tourist has
no idea about. For example, a tourist having interest in art history then the same tourist
might also have interest towards ancient monuments. This type of model can be derived
by the usage of intelligent ML algorithms, which processes the huge amount of tourist
data.
– Privacy and robustness: Privacy and robustness are a great challenges in case of TRS
domain. It is very essential to protect the privacy as the systems consist of large amounts
of personal information. It is also very important to protect the systems from different
kinds of attacks.

6 TRS using social media analytics

Social media platforms like Twitter and Facebook, and others, such as Flickr, can be used
to store, search and upload images or videos [78] (see Fig. 4). Foursquare App offers
personalized advice for locations near a current user location based on user background,
transactions and check-in background [107], etc. Few works have shown that the recom-
mendations can use Yahoo Maps, Google Maps and OpenstreetMaps, etc. [116]. Some other
researchers [78], have used by Wikipedia to extract POIs, latitude/longitude coordinates of
POIs and categories of user interest. Rashid et al. [113] present a thorough literature assess-
ment based on 109 high-quality research publications selected after a rigorous screening
approach. According to the statistics, there are eight basic categories of social media inten-
tion. A taxonomy of the tactics and approaches used for intention mining is also included
in this work. The social media intention may be used to determine tourist intentions, trans-
portation plans, and involvement in online travel businesses. This can further be used to
create a tourist recommendation system.
TRS requires two kinds of information namely, information about users’ preferences
and tastes and information about different products. Both can be obtained from the social
media. Social media may be used to detect new products or to apply more characteristics
to existing product in the social media database. Moreover, the social media concen-
trates on the identification of behavioral patterns, desires, and personal characteristics of
users.
There are few techniques which are used in the recommendation process from RS such
as CBRS, CFRS, KBRS, HRS etc. But due to the rapid development in the tourism industry,
these techniques are not adequate to provide recommendations. Therefore, a few techniques
have been specifically designed for the TRS domain. A few techniques have also been
developed using AI techniques to provide better recommendation.
8996 Multimedia Tools and Applications (2023) 82:8983–9027

Fig. 4 TRS using social media analytics

7 Techniques available in TRS

TRS techniques can be categorized in following ways. First, some of the RS techniques
which are used to design TRS have been highlighted. Then the techniques which are spe-
cially designed for TRS domain have been presented. Finally, the TRS techniques which
are developed based on AI techniques have been explained.

7.1 TRS techniques from RS

Garcia-Crespo et al. [44] discussed about Sem-Fit which is a semantic TRS, recommending
hotels to tourists. It helps to reduce the cost incurred by tourists in order to search for
accommodations, their location, and other amenities based on the taste of the tourist. It uses
fuzzy logic and ontologies of the hotel industry domain. Linaza et al. [80] presents a TRS
which generates personalized trips for the tourist along with the images of the attractions.
Details of the tourist are captured and used by TRS to present the list of tourist locations as
per his interest.
Lorenzi et al. [85] presents a TRS, named PersonalTour which suggests best travel
packages to the tourist, and is based on the concept of Distributed AI wherein a single rec-
ommendation request is split into partial recommendations handled by different individual
agents. Montejo-Raez et al. [101] presents Otium, which is a planning system that helps
in scheduling the tasks of a tourist by creating an agenda within the tour dates by con-
sidering the user’s taste and other parameters. In Fig. 5, there are two different kinds of
databases, namely, profiles database which is used for storing the user preferences and event
database which is responsible for storing various events such as sports show, dance show,
etc., The web interface which is Ajax based, helps to increase the system performance. Ulti-
mately, event database is updated using an off-line process which gathers the various events
Multimedia Tools and Applications (2023) 82:8983–9027 8997

Fig. 5 Otium Architecture [102]

using web services. Figure 6 represents the CORS which consist of three main components,
namely Recommender, Constraint Generator and Constraint Solver. A Recommender takes
input from the user’s rated component. A Constraint generator takes the constraints such as
travel distance, budget, etc, as input. A Constraint solver works as an optimizer of both rec-
ommender and constraint generator. The recommendation component takes input such as
trip to a country, tourist attraction, etc, and produces a suitable configuration. An example
of KBRS system which is used in TRS has been presented in Fig. 7. In Fig. 7, the client
requests the server for various information such POI, maps, etc., which depends on various
factors such as the actual geo-coordinates, trip destination,and preferences of the user. The
user interface provides a huge label of interaction with the users. The planning engine com-
ponent is used to generate the plan. Map Server (MS) acquires the map from external server
and subsequently, the route server calculates the route from the Google map. Ge-Ontology

Fig. 6 Complex Objects Recommender System (CORS) [121]

8998 Multimedia Tools and Applications (2023) 82:8983–9027

Fig. 7 GeOasis architecture [92]

is basically codification of the domain knowledge. The Web Map Server (WMS) protocol
serves the geo-referenced map images which are generated by a MS. Figure 8 is an example
of HRS system which is used in TRS. In Fig. 8, the engine is connected with the RS and the
e-Tourism control subsystem which translates the recommendation query of the user into a
group of understandable data [123]. On the contrary, it also translate the list of recommen-
dation which are then sent across to the Control System (CS). The recommendation process
is managed by the Control RS which collects recommendation list from the RS modules.
The HRS then forwards these lists and adds the other data for generating the recommended
items set [123].
Additionally, there are many techniques which are used to improve the accuracy of TRS
by applying the associative words of data-mining, AI and agent etc.

Fig. 8 GRSK Architecture [123]

Multimedia Tools and Applications (2023) 82:8983–9027 8999

7.2 Techniques available in TRS domain

Some researchers are focusing on developing different techniques which are associated with
the traveler behavior (see Tables 3, 4, 5 and 6). For instance, MinFeng et al. [99], developed
an isotropic dynamic spatial lag panel origin eDestination (OD) travel flow model, in order
to comprehend the travel pattern of the Australian domestic and international travellers. In
spite of these works, there are few challenges which need to be understood and captured
fully. Few of the points regarding the behavior of international travelers are highlighted
below.
1) Capturing of tourist data: Survey and opinion poll constitute the most common mecha-
nisms for understanding the travel behavior of tourist. However these methods are time
consuming and many of the tourist may not respond and also, the data might not reflect
the actual travel pattern
2) Travel Preferences: There is a common assumption that the different tourist group
exhibits different travel behaviors [78]. Tourist belonging to different countries may
have different preferences such as their staying period, types of travel attractions, etc.
Li et al. [72] provide a consistent technique for evaluating context-aware similarities
across distinct users by examining the influence of temporal and geographical aspects
on the users. They offer a method for creating dynamically unique POI suggestion
lists for each user based on the user’s current context information. Sarkar et al. [120]
provided a TRS framework for recommending multiple itineraries. The authors have
also presented a framework in which multiple itineraries were suggested based on POI
within POI [119].
Recently the advantages of mobile technologies, multimedia [146] and also the popular-
ity of social media have boosted the growth of huge amount of data generated by the user
such as photographs of different places with built-in GPS technology. The exact geographi-
cal location of the traveler can be obtained from the latitude, longitude, and other metadata
of each of the photograph. The geo-tagged data helps to extract the necessary information
such as the tourists’ travel behavior, travel pattern, tag completion, POI extraction, etc. Yin
et al. [152] described an approach for mining the trajectory patterns which are frequently
visited using the geo-tagged collection of different pictures. The authors suggested that the
GPS locations need to be identified before extracting trajectory pattern.
In a single day, the trajectories of a particular tourist are created depending on his/her
visit to the different POI. If a trajectory pattern is repeated frequently, it is considered as
a pattern of frequent trajectory. Geotagging includes attaching the GPS coordinates to the
photographs using the Exchangeable Image File Format (EXIF) [35]. During the process
of Reverse geocoding, the name of the place from the photograph. A reference database
based on geospatial data is necessary to extract different parameters. This process is useful
to identify the user’s POI but may not be sufficient enough to mine the necessary data [35].
Liu et al. [81] proposed the geo-tagged photo data set based approach which divides the
classification process into three parts namely, city, important places and tourist behavior.
Ickjai et al. [56] proposed a mining framework for POI, taking the Flicker data set into con-
sideration. The data is then further preprocessed in order to ensure that the duplicates data
does not exist and clustering algorithm is subsequently applied. The multiple photographs
taken by user at a particular location are considered for the purpose of a single case study.
Furthermore, to identify the clustered POI pattern, the preprocessed datasets are fed into the
clustering algorithm.
9000 Multimedia Tools and Applications (2023) 82:8983–9027

Table 3 Summary of Techniques which are available in TRS domain

Authors Summery Disadvantages

[115] – Successive time-stamp for recommendation This work does not truly reveal the real-time
of the attraction. situations of a particular attraction.
– Personalized recommendation problem was
resolved using the factorizing personalized
Markov chain (FBMC) method.
[164] – Spatial-temporal latent ranking model to This work does not take into account the
recommend the best possible successive personalized factor and also does not truly
attractions considering the time significant. reveal the real-time situations of a particular
attraction.
[73] – time-aware factorizing personalized Markov Complex model wherein every calculation
chain (TA-FBMC) model. uses all the check-ins data. Also, the authors
– 4-dimension static tensor in order to store have taken the user as an extra dimension
both the temporal and spatial check-ins data. which is Redundant. Hence, in case a new
– Time-decay factor by calculating the difference user is included in the RS the performance
between two successive check-ins. is degraded.
[74] – Completion algorithm based on fast low-rank There exists an issue wherein the
dynamic tensor. recommendation is not matching to a
– Recommendation using the pigeon-inspired particular tourist as the data is taken from
heuristic algorithm. the users who do not belong to that particular
location. This problem exists when there are
a large number of outsider in a particular place.
[159] – Generate the entire package for the tourist A specific POI has not been taken into account.
which contains the attractions and the optimum
sequence in which the tourist needs to
visit them.
– Uses the collaborative filtering technique to
determine the preferred attraction of the
particular tourist.
[151] – Used Spatial-Temporal LDA (ST-LDA) to In case the tourist are on the same social
come up with an efficient attribute pruning networks there interest will be same.
algorithm.

After the clustering process has been completed on the data of the tourist attractions,
the identified clusters can be assumed to be as POI. In practice, a single cluster is formed
using the geo-tagged photographs of several POI. Consequently, there will be many POI
which may not have been mined. To resolve this issue,T-h et al. [134] came up with a
novel framework for POI mining which is based on clustering at two levels namely, random
walk and constrained clustering. The authors had separated out large amount of photos
which were geo-tagged into several groups in the case of random walk clustering. On the
contrary, constrained clustering divides the clusters which are associated with several POIs
into sub-groups.
Table 4 Summary of Techniques which are available in TRS domain

Authors Summery Disadvantages

[2] – Probabilistic model in order to established the relationship between The link between a particular season of the year and the specific attraction has not been
user-annotated tag and the attraction like-ability. done appropriately.
– Scores are computed from multiple LBSNs and integrated
using learning to rank techniques.
[152] – Mining the trajectory patterns which are frequently visited using The POI recommendation made are not at all diverse in nature.
the geo-tagged collection of different pictures.
– In a single day, the trajectories of a particular tourist are created
depending on his/her visit to the different POI.
[56] – Mining framework for POI, taking the Flicker data set into consideration. Many of the attractions are not mined using this particular technique.
Multimedia Tools and Applications (2023) 82:8983–9027

– To identify the clustered POI pattern, the preprocessed datasets are fed
into the clustering algorithm.
[160] – Manual investigation on real-world datasets The link between a particular season of the year and the specific attraction has not
been done appropriately. It does not recommend the attractions which are time-based
by taking into account the duration of the user’s visit to the specific POI.
– Co-pairwise ranking model by assuming that the POIs which are near
to the user’s previously rated POI are given a higher rank.
[90] – Understanding the different context such as time, date, weather, The recommendation does not take into account the similarity of users, but on the
collective wisdom of users, etc. contrary, uses a simple similarity computation to make the recommendation.
[28] – Guessed the favorite locations of the users from Flickr photos by It does not recommend the attractions which are time-based by taking into account
measuring the similarity between the different tourists on the basis of the duration of the user’s visit to the specific POI.
travel histories and similarity values.
9001
 9002

Table 5 Summary of Techniques which are available in TRS domain

Authors Summery Disadvantages

[76] – Developed PERSTOUR algorithm to recommend tours which are The exact calculations of the arrival and departure time is not possible
personalized on the basis of popularity of the POI and preferences through the geo-tagged photograph that the tourist uploads.
of the user.
[65] – Recommends a travel route for the tourist based on the geo-tagged It does not recommend the attractions which are time-based by taking
and time-stamped photographs held by Flickr. into account the duration of the user’s visit to the to the specific POI.
– Depending upon the photographer’s behavior, a photographer
behavior model has been developed.
[77] – Recommends a travel route for the tourist based on the geo-tagged The exact calculations of the arrival and departure time is not possible
and time-stamped photographs held by Flickr. through the geo-tagged photograph that the tourist uploads.
– Suggested a solution on the basis of clustering for a group of
members, for recommending POI and finally, for assigning guides
using the technique of integer programming.
[64] – Recommend a tourist based on data from the internet sources like During the tourist’s visit to several attractions the context information
social media and various other websites. is not considered.
– Used the technique of Collaborative filtering in order to determine
few of the attraction.
[22] – Used the Chebyshev polynomial approximation method to derive The link between a particular season of the year and the specific
the tourist preferences based on the location and details present for attraction has not been done appropriately.
that particular location.
[153] – Time-aware metric embedding algorithm wherein the factor of the The link between a particular season of the year and the specific
user visiting the attractions at different point of time is taken attraction has not been done appropriately.
into account.
– Dynamic approach.
Multimedia Tools and Applications (2023) 82:8983–9027
Table 6 Summary of Techniques which are available in TRS domain

Authors Summery Disadvantages

[14] – Personalized Touristic Tour (PTT) framework and modeled after It is not possible to calculate the transit time between the different
the Generalized Maximum Coverage problem. attractions visited by the tourist from the geo-tagged photo.
– Unsupervised method for ex- tracking the tourists movement
Multimedia Tools and Applications (2023) 82:8983–9027

pattern in a particular area.

[86] – Helps to suggest the best possible trip within the constraint of The link between a particular season of the year and the specific
the travel time of the tourist. attraction has not been done appropriately.
[34] – Explained different methods for personalized POI recommendation The specific attraction has not been identified appropriately with
such as computing algorithm of the similarity, the first POI selection respect to the particular time duration (e.g, duration of a winter
and final POI recommendation method. or summer season).
9003
9004 Multimedia Tools and Applications (2023) 82:8983–9027

Sieh et al. [129] proposed a time-sensitive route RS named as TripRouter by considering

the popularity of places, order of visiting of different places with time and also the proper
timings for transition from one place to another. The authors have considered two major
functions as follows.
– Developing the route using the source location of the user and the starting time.
– Compose the path at a particular starting time between source and the target location.
Yuan et al. [160] conducted a manual investigation on real-world datasets and found that,
around multiple centers, each of the traits is likely to cluster around multiple centres. As a
consequence, the authors have proposed co-pairwise ranking model by assuming that the
POIs which are near to the user’s previously rated POI are given a higher rank. The proposed
method can have a greater liking for one alternative place over another from non-observed
rating pairs, and also tends to alleviate the problem of sparsity of matrix factorization.
Majid et al. [90] devised a system that mines the POI of the tourists and sequences of
travel from the geo-tagged photos by understanding the different context such as time, date,
weather, collective wisdom of users, etc. De C. et al. [28] extracted the Flickr photographs
and accumulated them into a graph and subsequently given different types of itineraries
based on the availability of the users. Cheng et al. [23] came up with a RS algorithm based
on the particular profiles of the user such as age, gender etc. Clements et al. [24] guessed
the favorite locations of the users from Flickr photos by measuring the similarity between
the different tourists on the basis of travel histories and similarity values.
The geo-tagged social media data helps in identifying the various tourist’s landmarks as
well as the tourist hotspots using various clustering methods. Brilhante et al. [14], [15] pro-
posed TripBuilder, which is a Personalized Touristic Tour (PTT) framework and modeled
after the Generalized Maximum Coverage problem. The authors have considered Flicker
for data of the original itineraries of the tourists and matched them with the touristic POI
from Wikipedia. Also, the authors have introduced an unsupervised method for extracting
the tourists movement pattern in a particular area. This method stores the tourist’s POI, pop-
ularity, visiting times, etc. Lu et al. [86] proposed Trip-Mine which is developed based on
the data mining based approach. The work helps to suggest the best possible trip within the
constraint of the travel time of the tourist. The authors have claimed that this works is the
first work which helps to make an efficient plan within the travel time constraints.
Souffriauet al. [132] came up with an approach which helps the tourist in taking fast
decision using the mobile device. The authors have solved the Tourist Trip Design Problems
(TTDP), which takes into account a set of location point having a particular score and the
aim is to increase the cumulative score of the location where the user has visited. On the
basis of the vector space model, the scores are calculated which represents the interests
about the particular location.
Vansteenwegen et al. [140] proposed an approach for Mobile Tourist Guide (MTG)
which acts as a user-friendly guide for the tourist. The mobile device can immediately rec-
ommend a holiday plan on the basis of real-time data. The authors have considered the
TTDP problem as an extension to the Orienteering problem. E-y et al. [34] have explained
different methods for personalized POI recommendation such as Computing Algorithm
of the Similarity, the first POI selection and final POI recommendation method. Lim
et al. [76] developed a model of tourist interests over time by considering his past visit to the
POI and the geo-tagged photographs. The author has developed PERSTOUR algorithm to
Multimedia Tools and Applications (2023) 82:8983–9027 9005

recommend tours which are personalized on the basis of popularity of the POI and pref-
erences of the user. The following steps have been suggested for providing the best
recommendation to the tourist.
– Initially, the POI visits determined based on the geo-tagged photos.
– A travel history is constructed by storing the tourists’ POI visits list.
– Recommendation is based on the PERSTOUR.
Kurashima et al. [65] came up with a technique which recommends a travel route for the
tourist based on the geo-tagged and time-stamped photographs held by Flickr. Depend-
ing upon the photographer’s behavior, a photographer behavior model has been developed,
which consist of two models namely, topic model which is responsible for calculating the
user’s individual preferences and the Markov model which is responsible for finding out the
routes of the photographers. Lime al [79] introduced a problem GROUPTOURREC which
addressed many challenges such as forming a group of members who possess similar inter-
ests, recommending POI. In order to resolve this problem, the authors suggested a solution
on the basis of clustering for a group of members, for recommending POI and finally, for
assigning guides using the technique of integer programming.
Maruyama et al. [94] proposed p-Tour, an individual navigation device which calculates
the route of the tourist. The author, have used a variant of the TSP with profits which focuses
on determining a circuit which will reduce the cost after subtraction of the profit collected.
Yu et al. [158] tried to solve a personalized tour recommendation travel package in order
to predict an attractive tourist plan. The authors, have considered three problem statements,
like which place should I go? How long can I stay in a tourist place?, and what should
be the appropriate sequence of visiting the places?. Authors have modeled user references,
location popularity, and location dependence. To discover the user’s POI, the authors, have
divided the total duration of travel into small time intervals and select the POI’s from a set
of POI locations for planning the route and compute the rank on the basis of user pref-
erences. Zheng et al. [165] came up with two types of recommendation engine based on
mining the GPS traces of different users. The first approach recommends based on the loca-
tions which are on top of the list of interests of the particular user, by modeling a Tree-based
Hierarchical Graph (TBHG), consist of user’s location history. The second one is a per-
sonalized recommendation which provides a personal with matching of their locations her
travel preferences.
Yuan et al. [161] suggested a framework named as DRoF which helps to discover the
different regions of a different functions in a city by considering mobility of the users in
that region and their POI. The author, have divided the city into different segments such as
highway and urban express way, etc. and using a topic-based inference model, authors have
inferred the functions (POI’s categories, restaurants, etc. ) of each region. Uddin et al. [137]
showed how to measure the Region of Interests (ROIs) in databases. ROIs defined as the
region where the number of moving objects are present for a given time interval. The pro-
posed approach does not depend on the ROIs parameter, which helps to identify RoIs very
fast without rescanning the entire database. Yoon et al. [154] proposed recommendation
techniques for large itineraries which are efficient and balanced using multiple GPS tra-
jectories. To receive an itinerary recommendation, user needs to generate a small query, a
finishing point and, also the travel duration. Lu et al. [88] tried to solved automatic travel
route planning problem for tourists. The authors collected around 20 million geo-tagged
photographs from www.panoramio.com, which helped in coming up with a route plan
based on the user’s preferences. Liu et al. [82] analyzed the different characteristics of the
9006 Multimedia Tools and Applications (2023) 82:8983–9027

various travel packages and developed a model, named Tourist-Area-Season Topic (TAST),
which extracts different topics based on the intrinsic features of the landscapes. Clements
et al. [24] came up with a model to recommend tourist location using the tourist’s visit-
ing history. De et al. [28] developed an approach which automatically constructs intra-city
travel itineraries based on geo-temporal breadcrumbs which are left by the tourists.
Chih et al. [135] came up with a personalized RS which helps to provide suggestions
about a sequence of visiting spots when the users takes the photo of a current scenic spot.
The authors have used data mining techniques to extract and distinguish the various users’
preferences. Ashbrook et al. [5] discussed about location context which helps to develop
a predictive model of the tourist’s future movements. The authors, have proposed a sys-
tem which automatically clusters the user’s GPS data which can be taken over an extended
period of time. Liliana et al. [75] proposed INTRIGUE which is a prototype tourist infor-
mation server. It presents the necessary information about different areas of Torino city.
INTRIGUE helps to recommend sightseeing destinations as well as itineraries keeping
in mind the preferences of heterogeneous group of tourists and group member’s require-
ments. Maruyama et al. [94] came up with a Personal Navigation System (PNS) which
helps a tourist to navigate through multiple destinations. PNS computes the route including
user’s destination by satisfying the requirements of the tourists and navigates the user. Wu
et al. [145] formulated the problem for scheduling the plan based on the weather condition.
The authors, have composed schedules for changing the weather probabilistically when the
future weather probability has been given. Hu et al. [52] proposed a Stamina-aware Sight-
seeing Tour Problem (SASTP), which says that, a tourist can propose various sightseeing
spots to be covered in a certain time interval and stamina budget. The spot is visited using
different methods such as short and guided tour etc., and the method needs some quantity
of stamina and time and gives a certain level of satisfaction to the tourist. Yin et al. [151],
have used Spatial-Temporal LDA (ST-LDA) to come up with an efficient attribute pruning
algorithm which is fast to recommend the attraction for the tourist.
Kotiloglu et al. [64], have proposed an algorithm that recommend a tourist based on data
from the internet sources like social media and various other websites. The authors have
used the technique of Collaborative filtering in order to determine few of the attraction that
will provide maximum satisfaction to a particular tourist. Chen et al. [22], have used the
Chebyshev polynomial approximation method to derive the tourist preferences based on the
location and details present for that particular location.
Ying et al. [153] suggested a time-aware metric algorithm that takes into account the
user’s visiting time at an attraction. In addition, the effect was viewed over a span of
time, making this strategy complex. In order to determine the relationship between the user
annotations tag and an attraction resemblance, Aliannejadi et al. [2] established a proba-
ble model. The results were measured from various LBSN resources and combined with
learning to rank strategies. Taking into account the significant time, Zhao et al. [164] sug-
gested a spatial-temporal latent model for suggesting best possible successive attractions. Li
et al. [74] suggested to use the 4-dimensional static tensor to store all the data from time and
space inspection for the individual visitor, using the Time-aware Factorizing Personalized
Markov Chain (TA-FPMC). By measuring the difference between two consecutive tests, the
authors took into account the time-decay element. The model was extremely dynamic and
all estimates were based on all check-in knowledge. In that work, authors often treated the
consumer as a needless extra layer. Therefore, the efficiency of a potential consumer would
become diminished if it is used in the recommendation Framework.
The classification of itineraries plays an essential role in understanding the nature of
destination. Dietz et al. [30] have proposed a crowd sourcing based wondary framework
Multimedia Tools and Applications (2023) 82:8983–9027 9007

which is described to schedule and communicate the trips globally. The authors represented
an innovative framework such as a content oriented itinerary recommendation to bypass the
item detection and drive off problem of the trip model.
Ayala et al. [6] proposed three new approaches such as PHILS, TRILS and SILS to solve
the problem of transportation from one POI to other POIs which is not considered by the
existing models. The mentioned techniques develop the visit schedules and fine tune it based
on the available data. The author did an experiment on Izmir city transport details and its
POI data set and found that their recommendation techniques provide dynamic visit sched-
ule with real time data and outperforms baseline algorithms. The average travel duration
helps for effective processing and the solution was adjusting the visit schedule appropri-
ately [69]. Laβ et al. [69] used different measurement approaches for different schedules.
To provide a real value based optimal visit solution along with travel time was the main
objective of this work. Only few Internet oriented mobile applications are available for tour
trip recommendation of POIs. Already seen POIs, moment of visit, open and closing times
of POIs and climate were the main focus of this work.
Lee et al. [70] discussed how the online intelligent tourism data can be helpful for the
tourists as well as South Korea’s tourism business. Based on big network data analytics, this
study concluded that the offline aspirant community had more tourism information rather
than online community.
Khatibi et al. [62] studied that the proper forecast on the upcoming incidents is very
much needed for the tourism business to attract the tourists and be ready for satisfying
them. Many business profits are mainly dependent on the tourism industry capability and
its proper predictions of future requirement in every field of the country. Also, this study
discussed that, how the social media, tourism website and natural data can help to predict
to attract the local and global tourists. Li et al. [72] discussed that due to the Internet, the
‘Top-K’ best places to visit (BP2V) suggestions are the primary yield factor for online travel
groups. However, they need to give more attention to what factors are mainly contributing
to finding the ‘Top-K’ (BP2V). Also had discussed the other related factors like purchasing,
foodstuff and enjoyment kinds of characteristics of the POIs.

7.3 Summery and insight

This work also surveys a few works that are specially intended for the TRS domain. Some
researchers are concentrating on creating distinct methods that are related to the conduct of
travellers. The geo-tagged data is used to extract the tourist information needed. Geo-tagged
data enables to acquire the necessary tourist information. Therefore, numerous works have
been found, based on geo-tagged data. Few works have been used with a 4-dimensional
static tensor to store both time and space check-in information for a specific tourist. How-
ever, the efficiency will be degraded if a new user is included in the RS. Therefore, a quick
low-level dynamic tensor is also used by the researchers. Itinerary classification plays a
key role in knowing the destination’s nature. There are few works based on the planning
of the itinerary. The researchers are also using AI methods given in Table 12 to solve the
tourism-related problem.

7.4 TRS developed based on the AI techniques

There are few techniques available for solving the issues of TRS based on the AI techniques
(see Tables 7, 8, 9, 10 and 11). The techniques are given below (Table 12).
 9008

Table 7 Summary of Multi-agent systems

Multi-agent systems Features System requirements

MARec approach [38] – Recommendations Through Expert and Trusted Agents Random and WTrust
SAMAP [20] – Integrates the module which can dynamically capture the models of user FF, IPSS, SIADEX
– ability to generate attractive tour plans based on the past experiences of the user
PaTac [20] – Provide services ubiquitously to the citizens and tourist RFID, content-based filtering method
TCT ontological recommendation [18] – A personalized travel route is recommended by the agents ACO, Fuzzy Inference Engine, CDA, TRRA, LTM
Itchy Feet [127] – Electronic Institutions for regulating the environment also to provide a GUI, 3-D graphics, On-line Communities
framework which has been used by Software
– 3D virtual world generates a lively community Agent
e-Tourism MAS Architecture [125] – Used genetic algorithm Tourism ontology
– Used the problem of traveling salesman
PersonalTour [85] – Agents acts as an expert Collaboration of multiple agents
– The knowledge about the travel services and stored into its knowledge base
and helps to produce an attractive recommendation
Multiagent TRS [19] – Travel Assistant Agent (T-Agent ) developed by graded BDI agent g-BDI agent
MARST [9] – Reputation Based Collaborative Filtering (RbCF) approach JSP, JADE, MySql 5.0.24
Multimedia Tools and Applications (2023) 82:8983–9027
Table 8 Summary of optimization techniques

Optimization techniques Features System requirements

TCT ontology recommendation [18] – Route generation Ant Colony Optimization

– Route customization
PET [40] – Routing system for the generation of tourist routes that are personalized MySql, Apache Tomcat, Google Web Toolkit (GWT)
The City Trip Planner [141] – Routes for five cities in Belgium Apache web server, JAVA Script, PHP
– Greedy Randomized Adaptive Search Procedure (GRASP)
Multimedia Tools and Applications (2023) 82:8983–9027

Context aware tourist app [95] – Taking a decision about users attraction bssed on the relevant contextual Fuzzy logic, principal component analysis, and
information Artificial neural networks,
PSOiA [26] – Particles are located randomly C++, GNU GCC, Linux
– Particle local and global best position identified
TTDP model [31] – Helps to plan the route in a new TRS System Visual Studio 2015
Managing travel demand [71] – Individual decision for traveling Neural Network approach
– By using the machine learning technique infer unobserved preferences
from mobile records
9009
Table 9 Summary of automatic clustering
9010

Automatic clustering Features System requirements

SAMAP [20] – Integrates the module which can dynamically capture the models of user FF, IPSS, SIADEX
– Past experiences of the same users and ability to generates attractive plans
Ontology-based feature extraction – k-means algorithm for identify the tourist segments set, to identifying the SWOOGLE search engine,
method [142] users which have similar demographic characteristics Tourism ontology
– Identifying the users who have provided their individual ratings
OWA [148] – Aggregating criteria guided by quantifiers fuzzy logic
Hybrid context aware system [36] – recommendation system uses the fuzzy c- means as well as k-means Fuzzy data analysis
uncertain version
– each cluster and for each object, the membership lies between the value of 0
and 1. For each group the summation of membership degrees equals 1
A hybrid recommendation approach [89] – Classification based on association is applied Fuzzy c-means
Tourist@[8] – Clustering has been applied all the time that 10 users can join newly Content-based and Collaborative
recommendation strategies
– periodically it is require to recompute the classes
ClusDM[139] – Solves multi-criteria decision problems Lisp and Java
SPETA [43] – Uses the knowledge of the user’s present location, preferences and GIS modul
the user’s past locations.
– suggested a TRS by using the Support Vector Machines (SVMs)
where the different user preferences are stored into the vector.
W2Go[39] – Travel based on the landmark ranking Normalized Cuts algorithm, Flickr API
– Developed using the statistics of photo from Flickr and also the
knowledge of the user from the Yahoo travel guide
Community-based Social Tag – Automatic clustering of Social Tags (ST) with the help of Seekda search engine, Web Services
Clustering [108] community detection. Description Language (WSDL)
– Identified that the ST network has scale in free properties
Multimedia Tools and Applications (2023) 82:8983–9027
Table 10 Summary of Management of Uncertainty

Management of Uncertainty Features System requirements

ITAS [53] – Possible number of attributes influence directly on the probability Windows 2008 Server IIS7.0, Visual Studio 2008, Netica
where the users have an certain tourist point of interest 4.09, Google Maps API
– Integration of EBM model, Bayesian network
Online personalized RS which is enabled – linkage about the users when users linking a certain tourist destination Java/Servlet, JSP, ArcWeb servic, Netica-J 2.21
Multimedia Tools and Applications (2023) 82:8983–9027

semantically for planning the tour [53]

– Heterogeneous integration, personalized recommendation, adaptive
capability, spatial functions capability
Context aware tourist app [96] – Taking a decision about users attraction based on the relevant Fuzzy logic, principal component analysis, and Artificial
contextual information neural networks
Visitor Information System [67] – rule-based approach wherein a system has certain rules which Mobile services, Contextual Computing
generate recommended events based on the user preferences.
9011
 9012

Table 11 Summary of Knowledge representation

Knowledge Representation Features System requirements

3D-GIS Hybrid RS [124] – An interface based on 3D map and real-time location-aware recommendation GPRS and UMTS, C++, GNU/Linux
SMARTMUSEUM [118] – Ubiquitous RS for the Web of Data Kappa statistics
– Bridges the semantic gap that exists in the sensor inputs, heterogeneous
content descriptions and the user profiles. Ontologies are used in this case.
SigTur/E-Destination [13] – User preferences are stored into each node of the ontology Collaborative filtering
Ontology Population [117] – minimizes the cost for constructing the ontology Named entity recognition, Ontology Population phase
Ontology-based Framework [27] – Specific types of services are accessed Web Server 9.0, JSP, Enterprise Java Beans (EJB)
– Semantically enabled e-Tourism Portal
OAISIS [45] – collects data on the basis of semantic ontologies DataLift, CSV, JSON
Multimedia Tools and Applications (2023) 82:8983–9027
Multimedia Tools and Applications (2023) 82:8983–9027 9013

Table 12 Summary of recommendation methods using RS, TRS and AI Techniques

7.4.1 Multi-agent systems (MAS)

An agent is basically a computer system which has the ability to take an independent actions
on behalf of its particular owner or user [18]. On the other hand in case of MAS, multi-
ple agents work together for sharing their information, their activities, resources etc. and
producing a joint efficient outcome of a distributed problem (see Fig. 5).
Casali et al. [19] proposed a TRS which consists of multi-agent architecture where the
Travel Assistant Agent (T-Agent) is developed by graded BDI agent (see Fig. 9). Castillo
et al. [20] proposed a software tool named SAMAP which basically integrates the module
which can dynamically capture the models of the user, list of activities and subsequently

Fig. 9 Functional, MAS architecture [21]

9014 Multimedia Tools and Applications (2023) 82:8983–9027

generate attractive plans for the user. SAMAP is basically based on MAS and consist-
ing of user modeling, Case-Based Reasoning (CBR), interface agent and planning agent.
Ceccaroni et al. [21] proposed a TRS named as PaTac is basically consists of three main
component, namely the Content Collector (CC), agents, and Content Deliverer (CD). The
agents are responsible for storing the necessary information in ontologies and the CD
provide the services to the other component. The PaTac technique consists of different
components like Communication Agent (CA), Session Handling Agent (SHA), Profile
Managing Agent (PMA), Personalization Agent (PA), and Information Service Agent (ISA).
Initially, CA receives the query, encloses the Agent Communication Language (ACL) mes-
sage and subsequently submits to the SHA. The SHA is responsible for handling the user
session and through the PMA it initializes the Up and finally submits the PA user request.
PA is responsible to gather the preferences of the users via PMA and it also gathers the
content from the database and ontologies via ISA. Ultimately, by using the personalization
algorithm PA generates the result for the particular user. C-s et al. [18] proposed a technique
for Tainan City Travel (TCT) (see Fig. 6). The authors have used the concept of Travel Route
Recommendation Agent (TRRA) and Context Decision Agent (CDA). Initially, a domain
expert produces the ontology of Tainan City travel. A Location Transfer Mechanism (LTM)
has been used, which depends on the information received from TCT and is used to generate
the location value, which is subsequently saved into the ontology. TCT ontology is retrieved
by CDA in order to determine the tourist matched concept and then computes the seman-
tic as well as the context relation for each instance. On the basis of the output of CDA, a
fuzzy inference engine is constructed by TRRA to infer the context information. Sebastia
et al. [125] proposed a multi-agent architecture for TRS. The authors have proposed a
system which integrates various agents, helps to dynamically collects the Up and also to
analyze the user activities with the help of experience from the user. Lorenzi et al. [85] came
up with a recommendation technique on the basis of multi-agent platform named as “Per-
sonalTour”. When a tourist request for recommendation, the agents acts as an expert and
provides an attractive recommendation. Lorenzi et al. [84] proposed a RS based on the the
trust and the expert agent. To bring down the communication load and to increase the quality
of information exchanged between the agents, authors have considered the communication
from the trusted sources. Seidel et al. [127] proposed a technique for TRS which is built on
the 3D e-tourism recommendation technique named as Itchy Feet. The authors have con-
sidered the Electronic Institutions for regulating the environment and have also provided a
framework which has been used by the Software Agent (SA). The SA provides information
from the environment where social interactions are implicitly addressed and the 3D virtual
world generates a lively community. Bedi et al. [9] suggested a TRS technique based on
the Reputation based Collaborative Filtering (RbCF) approach named as MARST. MARST
consist of various agents and has been described as follows:

– Specialized agents (Data Repository Agent (DRA)): the necessary information are
extracted or aggregated from the web services.
– Items Reputation Agent (IRA): calculates the reputation of all the items.
– Similarity Agent (SA): calculates the similarity between the users.
– System Agents (User Agent (UA)): helps to interact with other agents. UA receives the
user query request and transfers it to either the HRA, RRA or PRA.
– Hotel Recommendation Agent (HRA): it takes the query from the UA
– Place Recommendation Agent (PRA) and Restaurant Recommendation Agent (RRA):
Service recommendation is computed for both the cases.
Multimedia Tools and Applications (2023) 82:8983–9027 9015

7.4.2 Optimization techniques

There are several optimization techniques that are used for optimizing TRS. C-s et al. [18]
proposed Ant Colony Optimization (ACO) technique for finding the personalized travel
route (see Fig. 10). Garcia et al. [40] proposed a routing system for the generation of tourist
routes that are personalized. Vansteenwegen et al. [141] proposed an Expert System (ES)
based on the Greedy Randomized Adaptive Search Procedure (GRASP) algorithm. Mee-
han et al. [95] developed an intelligent RS by accessing the user’s context using Artificial
Neural Network (ANN). D-c et al. [25] used PSO-inspired Algorithm(PSOiA) for solving
the TOP problem. In PSOiA, at the very beginning, the particles are located randomly and
then iteratively location of each particle is replaced by new one using Randomized Heuristic
Algorithm (RHA). The local search algorithm is used to determine the local and global best
position for every single particle [26]. Ding et al. [31] proposed a TRS using route planning
based on Fireworks Algorithm [136]. The authors have considered the Tourist Trip Design
Problem (TTDP), which is basically a discrete optimization technique. Leng et al. [71] came
up with a technique which does location based recommendation depending on the Geo loca-
tion data captured by the mobile phone. The authors have used optimization technique for
improving the efficiency of a system.

7.4.3 Automatic clustering

Clustering is defined by the act of partitioning the dataset which is unlabeled into groups
of objects which are similar [133]. The objects of the same group are similar but dissim-
ilar with other group’s object. There are many recommendation systems which has been
developed using collaborative filtering, where the users are partitioned into groups and user
of each group share similar characteristics. On the basis of demographic information, user
individual ratings of each item and user preferences, the similarity of group users has been
identified. Castillo et al. [20] identified the similarity of users on the basis of a domain ontol-
ogy. Here the system concludes that, a user preferring Cinema is more similar to users who

Fig. 10 Ontological recommendation multi-agent Structure for Tainan City travel [18]
9016 Multimedia Tools and Applications (2023) 82:8983–9027

enjoy the Theatre than any Spots activities. Vicient et al. [142], used the k-means clustering
algorithm to identify different tourist segments having similar demographic characteristics
and individual ratings. Yager et al. [148], and Dujmovic et al. [32] used OWA and LSP
to add similarities based on users’ ratings using the well-known Pearson correlation as a
similarity measure.
Fenza et al. [36], proposed a RS using the fuzzy c-means as well as k-means uncertain
version. The authors have defined that for each cluster and for each object, the membership
lies between the values 0 and 1 and for each group the summation of the membership is
equals 1. Lucas et al. [89] proposed Personalized Sightseeing Information System (PSIS)
using same technique [36]. Batet et al. [8], proposed a system wherein the clusters are
formed every time when a group of ten new users join the system. The underlying principal
states that the addition of users requires the classes to be recomputed periodically. Valls
et al. [139] proposed a clustering system in which a hierarchy of different classes has been
created on the basis of the user’s interest.
Gao et al. [39] developed a system for the proper guidance of travel based on the land-
mark ranking. The system has been developed using the statistics of photo from Flickr and
also the knowledge of the user from the guide present at Yahoo! Travel. Pan et al. [108]
developed a system on automatic clustering of Social Tags (ST) with the help of commu-
nity detection. Authors have identified that the ST network has scale in free properties.
Garcia-Crespo et al. [43] suggested a TRS by using Support Vector Machines (SVMs)
where different user preferences are stored into the vector. To calculate the gap between the
preferences of the user and the recommendable items, authors have considered SVMs for
identifying the most appropriate ones.

7.4.4 Management of uncertainty

Developing an intelligent RS by fully understanding the requirements of the tourist properly

is a challenging task. The demand of each tourist may be different and sometimes almost
same. A lot of works has been done towards developing an understanding the tourist activ-
ities based on AI techniques. Pearl et al. [111] used Bayesian network which is based on
acyclic graphs, where the edges of the graph represents relationships of causality between
available nodes. Later, Hsu et al. [50] used Bayesian network where the possible number of
attributes influence directly on the probability where the users have an certain tourist point
of interest. Huang et al. [53] proposed a network, wherein the attributes are influenced by
the type of the user and also there exists linkage between the users and the tourist destina-
tion. To handle the uncertainty of tourist there are some works which have been developed
based on the Fuzzy logic. A series of linguistic labels have been taken by Fuzzy variable
and also each label is considered a Fuzzy set where the membership function lies between 0
and 1. There are many users whose preferences can be represented using the Fuzzy sets and
fuzzy reasoning to calculates the user’s preferences, [18, 43]. A Restaurant can be stated as
either good or bad and is represented by the value between 0 and 1 but it is also possible
that for further analysis good or bad comments may instead be given, 2013 [96]. Lamsfus
et al. [67, 68] developed a RS based on the rule-based approach where in a system there are
certain rules which generates recommended events based on user preferences.

7.4.5 Knowledge representation (KR)

The clear information about the domain is called as the domain knowledge. The main issue
is to gather full knowledge from the domain, because it comes from different sources and
Multimedia Tools and Applications (2023) 82:8983–9027 9017

also should be integrated. For designing a knowledge base, individual and the relationship
should be clearly represented. Knowledge is required to solve any problem as well as to
design the program. AI can be used in a satisfactory manner for this purpose. Presently,
the most common method for representation of the domain knowledge is through ontolo-
gies, which helps to describe clearly and the ideas about the domain. The ontologies store
necessary information about different features which are used for generating an intelligent
RS. Wang et al. [143] considers a user and travel ontology where tourists’ demographic
characteristics as well as their preferences are modeled. [20, 91], and Ruotsalo et al. [118],
Alonso et al. [3] have also used ontologies for designing the activities of different tourists.
The authors used ontology-based similarity for calculating similar activities of different
tourists and providing recommendation to the tourist based on the technique of collaborative
filtering.
Borràs et al. [13], Moreno et al. [103], Sebastia et al. [124]; [126]; Garcia et al. [42],
have also used ontologies in their system. The user preferences are stored in each node
of the ontology [13, 103] and the values are generated by the users who have filled the
questionnaire about their experiences and preferences. Borràs et al. [13] proposed ontology-
based system to dynamically manage the user preferences.
Castillo et al. [20] proposed PaTac which uses different ontologies of restaurants,
entertainment, sports etc., and to integrate them authors have used standard temporal
and geo-location ontologies given by the World Wide Web Consortium (W3C). Lamsfus
et al. [67] proposed a system called CONCERT, wherein all the travel context is stored in
an ontology network called Contology. Ruı́z-Martı́nez et al. [117], and Vicient et al. [142]
came up with a technique in order to minimize the cost for constructing the ontology. Xu-yin
et al. [147] suggested a new approach on intelligent RS based on the WebGIS by applying
the knowledge representation theories of AI. Daramola et al. [27] suggested a TRS frame-
work which is based on ontology. The framework is divided into three layers, client-tier
where specific types of services are accessed, middle layer is called Semantically enabled
e-Tourism Portal and third layer is the layer of ontology such as destination, travel, accom-
modation, Restaurants and event ontology respectively. The architecture of this framework
has been shown in Fig. 11. Umanets et al. [138] proposed a RS based on the user current
location named as SPETA. Gasmi et al. [45] proposed an approach for TRS named as OAI-
SIS. The authors have provide the solution of data management based on the CrowdSensing
technique.

7.5 Summery and insight

Most of the techniques designed for TRS are either a client-server or a stand-alone system
where there is no opportunity to exchange information among agents. Many works have
been developed on the basis of MAS to solve this problem. On the contrary, there are several
optimization techniques that are used for optimizing TRS. The authors have used optimiza-
tion technique for improving the efficiency of a system. There are many recommendation
systems that have been created using clustering techniques, where tourists are divided into
groups and tourists of each group share comparable features.
Figure 12 shows that most of the existing works are using RS methods to solve the
TRS problem followed by the techniques specially designed for TRS and AI methods. As
discussed in Section 3.1, there are various factors which are considered for recommendation
in TRS. In the literature authors have used these factors either singly or jointly. Figure 13
shows the distribution of existing works with respect to different factors. As discussed in
9018 Multimedia Tools and Applications (2023) 82:8983–9027

Fig. 11 Ontology-based Architecture Framework

Section 3.1 TRS suggests different features to the tourist. Figure 14 shows the distribution
of existing works based on the features they suggest.

8 Future Research directions

For developing an intelligent RS, there exists few research challenges which have been
described as follows.
– Currently most of the RS are focusing on the representational view of the context.
However, it is very important to know the various context knowledge such as partial
and static, partial and dynamic, latent knowledge of context and also unknown context.
In such cases an intelligent system is required which can work with all these factors
and provide the best RS to the tourist.

Fig. 12 Tourism Recommendation system: Based on RS, TRS and artificial intelligence
Multimedia Tools and Applications (2023) 82:8983–9027 9019

Fig. 13 Tourism Recommendation using tour interest, popularity, travelling time and traffic condition

– Most of the existing works mentioned by previous researchers are based on prefiltering
method , postfiltering method as well as contextual modeling method. However there is
a need to explore alternative methods in a comprehensive manner and also to identifying
the limitations of these methods.
– During the clustering process for a large scale implementation it may be difficult to
derive the satisfaction functions.
– Sometimes when an item is shown to the user, instead the user follows the other item
which clearly shows that the user has more interest in the other item than the current
item at hand. So by designing a proper user interface (UI) an individual satisfaction can
be increased.
– Most of the work considers the accurate profile of user’s individual preferences but
in real scenario the data are probabilistic and also most of the work tries to solve this

Fig. 14 Tourism Recommendation using different features

9020 Multimedia Tools and Applications (2023) 82:8983–9027

uncertainty based on the Baysian networks but it is required to focus more on the
average and plurality voting strategies.
– Privacy preserving is a great challenge to designing a RS system as it requires user data
to make individual recommendations.
– The POI sequence recommender is based on some criteria like time, individual priority
and budget and its score. The score is calculated based on individual priority, trip dura-
tion between POIs and POI popularity. Hence, there is a need for an efficient optimizing
approach which increases the score of the related criteria.
– The tourist preference may overlap with time. So an itinerary gets the most importance
rather tourist’s interest preferences.
– The future work has opportunities like to consider the date, time and climate of the
location for the best recommendation to satisfy the user.
– Also, the future work has the scope like merging of POI popularity and tourist interest
with mobile application and provide timely POIs recommendation.
– If a user does not have any idea about the target place, in such cases there should be a
certain degree of diversity amongst all the items which are there in the recommendation
list.
– Sometimes there are large number of items available for the users and in such cases a
user may give ratings to few items only.
– The mobility of users poses great challenge and the user may need recommendation at
the new locations. In such cases it is very important to know the places that the user has
already visited along with his current location.
– To avoid wasting and waiting time of single transport mode during shifting process, the
future research direction has a scope of various form of transport suggestion like flight,
taxi, bus, etc. in itinerary recommendation along with merging of diverse groups like
family, youngsters, seniors, etc.
– To design a dynamic itinerary recommendation approach is a great challenge which
can overcome the loss of POI coverage due to environment factors like rain, flood, hot
climate, etc. The algorithm may be designed to suggest alternative safe POIs to visit in
such kind of environment.
– Same type of political, tradition, language itinerary suggestion and low cost as well as
visa-free countries, itinerary recommendation based on tourists interest is also a good
choice of future research direction.
– To recommend new user itinerary based on the old user feedback such as their ful-
fillment/disappointment ratings, time spent, interest, etc. of POIs, will enhance the
itinerary recommendation process.

9 Summary and conclusions

This work discusses the salient points about TRS and its importance to the tourists to plan
their trip in the best possible manner. Tourist data can be extracted from various social media
platforms and can be used further in TRS. Specific techniques which are targeted towards
TRS have been discussed in detail which clearly show us that the tourism domain is unique
in nature and TRS needs to handle the various aspect like POI, Itinerary, etc., in a special
manner. Several examples of TRS have been highlighted along with their salient features.
Also, future research direction has been suggested to motivate the young researchers.
Multimedia Tools and Applications (2023) 82:8983–9027 9021

Declarations
Conflict of Interests The authors have declared that there is no conflict of interests exist.

References

1. Al-Shamri MYH, Bharadwaj KK (2008) Fuzzy-genetic approach to recommender systems based on a

novel hybrid user model. Expert Syst Appl 35:1386–1399
2. Aliannejadi M, Crestani F (2018) Personalized context-aware point of interest recommendation. ACM
Trans Inf Syst 36:1–45
3. Alonso K, Zorrilla M, Iñan H, Palau M, Confalonieri R, Vázquez-Salceda J, Calle J, Castro E (2012)
Ontology-based tourism for all recommender and information retrieval system for interactive com-
munity displays. In: 2012 8th International Conference on Information Science and Digital Content
Technology (ICIDT2012) (pp 650–655). volume 3
4. Amatriain X, Jaimes A, Oliver N, Pujol J (2011) Data mining methods for recommender systems. In:
Recommender Systems Handbook, US (pp 39–71)
5. Ashbrook D, Starner T (2003) Using gps to learn significant locations and predict movement across
multiple users. Pers Ubiquit Comput 7:275–286
6. Ayala VAA, Gülsen KC, Alzogbi A, Färber M, Muñiz M, Lausen G (2017) A delay-robust touristic
plan recommendation using real-world public transportation information. In: RecTour@RecSys
7. Banerjee S, Chis M, Dangayach G (2010) Developing an adaptive learning based tourism information
system using ant colony metaphor. Studies Comput Intell 273:59–77
8. Batet M, Moreno A, Sanchez D, Isern D, Valls A (2012) Turist@: agent-based personalised recommen-
dation of touristic activities. Expert Syst Appl 39:7319–7329
9. Bedi P, Agarwal SK, Jindal V (2014) Richa Marst: multi-agent recommender system for e-tourism
using reputation based collaborative filtering. In: Proceedings of the 9th International Workshop on
Databases in Networked Information Systems - Volume 8381 DNIS 2014. Springer-Verlag, Berlin,
Heidelberg, pp 189–201
10. Billsus D, Pazzani M (2000) User modeling for adaptive news access. User Model User-Adap Inter
10:147–180
11. Bin C, Tianlong G, Sun Y, Chang L (2019) A personalized poi route recommendation system based on
heterogeneous tourism data and sequential pattern mining. Multimed Tool Appl, 78
12. Borrás J, Moreno A, Valls A (2014) Intelligent tourism recommender systems: a survey. Expert Syst
Appl 41:7370–7389
13. Borràs J, Valls A, Moreno A, Isern D (2012) Ontology-based management of uncertain preferences in
user profiles. In: Greco S, Bouchon-Meunier B, Coletti G, Fedrizzi M, Matarazzo B, Yager RR (eds)
Advances in Computational Intelligence (pp 127–136) Berlin: Springer
14. Brilhante I, Macedo JA, Nardini FM, Perego R, Renso C (2013) Where shall we go today?: plan-
ning touristic tours with tripbuilder. In: Proceedings of the 22nd ACM international conference on
Information & Knowledge Management (pp 757–762)
15. Brilhante IR, Macedo JA, Nardini FM, Perego R, Renso C (2015) On planning sightseeing tours with
tripbuilder. Inform Process Manag 51:1–15
16. Brudy F, Ledo D, Greenberg S, Butz A (2014) Is anyone looking? mitigating shoulder surfing on
public displays through awareness and protection. In: Proceedings of The International Symposium on
Pervasive Displays PerDis ’14 (pp. 1:1–1:6). New York, NY, USA: ACM
17. Burke R (2002) Hybrid recommender systems: survey and experiments. User Model User-Adap Inter
12:331–370
18. C-s L, Y-c C, M-h W (2009) Ontological recommendation multi-agent for tainan city travel. Expert
Syst Appl 36:6740–6753
19. Casali A, Godo L, Sierra C (2008) A tourism recommender agent: from theory to practice. Intel Artif
40:23–38
20. Castillo L, Armengol E, Onaindı́a E, Sebastiá L, González-Boticario J, Rodrı́guez A, Fernández S,
Arias JD, Borrajo D (2008) samap: An user-oriented adaptive system for planning tourist visits. Expert
Syst Appl 34:1318–1332
21. Ceccaroni L, Codina V, Palau M, Pous M (2009) Patac: Urban, ubiquitous, personalized services for
citizens and tourists. In: Third International Conference on Digital Society (pp 7–12)
22. Chen B, Yu S, Tang J, He M, Zeng Y (2017) Using function approximation for personalized point-of-
interest recommendation. Expert Syst Appl 79:225–235
9022 Multimedia Tools and Applications (2023) 82:8983–9027

23. Cheng A.-J., Chen Y.-Y., Huang Y.-T., Hsu WH, Liao H.-Y. M. (2011) Personalized travel recommen-
dation by mining people attributes from community-contributed photos. In: Proceedings of the 19th
ACM International Conference on Multimedia MM ’11 (pp 83–92). New York, NY, USA: ACM
24. Clements M, Serdyukov P, de Vries AP, Reinders MJT (2011) Personalised travel recommendation
based on location co-occurrence. IEEE Trans Knowl Data Eng, abs/1106.5213, 1–30
25. D-C D, Guibadj RN, Moukrim A (2013) An effective pso-inspired algorithm for the team orienteering
problem. Eur J Oper Res 229:332–344
26. Dang D.-C., Guibadj RN, Moukrim A (2013) An effective pso-inspired algorithm for the team
orienteering problem. Eur J Oper Res 229:332–34
27. Daramola O, Adigun M, Ayo C (2009) Building an ontology-based framework for tourism recommen-
dation services. Inform Commun Technol Tourism, 135–147
28. De C FM, Amer-Yahia S, Golbandi N, Lempel R, Yu C (2014) Constructing travel itineraries from
tagged geo-temporal breadcrumbs. In: Proceedings of the 19th International Conference on World Wide
Web. 94
29. Dey AK, Abowd GD, Salber D (2001) A conceptual framework and a toolkit for supporting the rapid
prototyping of context-aware applications. Human-Computer Int 16:97–166
30. Dietz LW, Weimert A (2018) Recommending crowdsourced travels on wondary. In: Proceedings of the
Workshop on Recommenders in Tourism, RecTour 2018, co-located with the 12th ACM Conference on
Recommender Systems (RecSys 2018), Vancouver, Canada, October, 7:13–17
31. Ding H, Liangjun K, Geng Z (2016) Route planning in a new tourist recommender system: a fireworks
algorithm based approach. In: 2016 IEEE Congress on Evolutionary Computation (CEC) (pp 4022–
028)
32. Dujmovic J, Nagashima H (2006) Lsp method and its use for evaluation of java ides. Int J Approx
Reason 41:3–22
33. E-y K, Kim H, Cho J (2006) Personalization method for tourist point of interest (poi) recommendation.
Knowledge-Based Intelligent Information and Engineering Systems, 392–400
34. E-Y K, Kim K, Cho J (2006) Personalization method for tourist point of interest (poi) recommenda-
tion. In: International Conference on Knowledge-Based and Intelligent Information and Engineering
Systems (pp 392–400)
35. Ennis A, Chen L, Nugent CD, Ioannidis G, Stan A (2013) High-level geospatial information discovery
and fusion for geocoded multimedia. Int J Pervasive Comput Commun 9:367–382
36. Fenza G, Fischetti E, Furno D, Loia V (2011) A hybrid context aware system for tourist guidance based
on collaborative filtering. In: IEEE International Conference on Fuzzy Systems (pp 131–138)
37. G-r X, Lin C, Yang Q, Xi W, H-J Z, Yu Y, Chen Z (2005) Scalable collaborative filtering using
cluster-based smoothing. In: Proceedings of the 28th Annual International ACM SIGIR Conference on
Research and Development in Information Retrieval. ACM, Salvador, Brazil, pp 114–121
38. Gao R, Li Z, Du B, Li X, Song C, Chang J, Liu D (2018) STSCR: exploring Spatial-Temporal
Sequential Influence and Social Information for Location Recommendation. Neurocomputing
39. Gao Y, Tang J, Hong R, Dai Q, S CT, Jain R (2010) W2go: a travel guidance system by automatic
landmark ranking. In: Proceedings of the 18th ACM international conference on Multimedia (pp 123–
132)
40. Garcia A, Arbelaitz O (2010) Personalized tourist route generation. In: International Conference on
Web Engineering (pp 486–497)
41. Garcia I, Sebastia L (2014) A negotiation framework for heterogeneous group recommendation. Expert
Syst Appl 41:1245–1261
42. Garcia I, Sebastia L, Onaindia E (2011) On the design of individual and group recommender systems
for tourism. Expert Syst Appl 38:7683–7692
43. Garcia-Crespo A, Chamizo J, Rivera I, Mencke M, Colomo-Palacios R, Gomez-Berbis JM (2009)
Speta: Social pervasive e-tourism advisor. Telematics Inform 26:306–315
44. Garcia-Crespo A, Lopez-Cuadrado JL, Colomo-Palacios R, Gonzalez-Carrasco I, Ruiz-Mezcua B
(2011) Sem-fit: a semantic based expert system to provide recommendations in the tourism domain.
Expert Syst Appl 38:13310–13319
45. Gasmi A, Tamani N, Faucher C, Ghamri-Doudane Y (2016) Oaisis: an ontological-based approach
for interlinking crowdsensing information systems. In: International Conference on Systems, Man and
Cybernetics (pp 9–12). October, 3995-4000
46. Gavalas D, Kenteris M (2011) A web-based pervasive recommendation system for mobile tourist
guides. Personal Ubiquituous Comput 15:759–770
47. Gavalas D, Konstantopoulos C, Mastakas K, Pantziou G (2014) A survey on algorithmic approaches
for solving tourist trip design problems. J Heuristics 20:291–328
Multimedia Tools and Applications (2023) 82:8983–9027 9023

48. Gergely R (2013) Influence of social media on tourism - especially among students of the university of
debrecen. Appl Studies Agribusiness Commerce - APSTRACT. 7:105–112
49. Herzog D, Promponas-Kefalas N, Wörndl W (2018) Integrating public displays into tourist trip rec-
ommender systems. In: Proceedings of the Workshop on Recommenders in Tourism, RecTour 2018,
co-located with the 12th ACM Conference on Recommender Systems (RecSys 2018), Vancouver,
Canada, October, 7:18–22
50. Hsu FM, Lin YT, Ho TK (2012) Design and implementation of an intelligent recommendation system
for tourist attractions: the integration of ebm model, Bayesian networks and google maps. Expert Syst
Appl 39:3257–3264
51. Hsu S, Wen DM-H, Lin H-C, Lee C.-C., Lee C.-H. (2007) Aimed: a personalized tv recommendation
system. In: European Conference on Interactive Television (pp 166–174)
52. Hu O-WB, R RG (2013) A mixed integer model for the stamina-aware sightseeing tour problem. In:
Extended Abstracts of the 14th International Conference on Computer Aided Systems Theory (pp 200–
202)
53. Huang Y, Bian L (2009) A Bayesian network and analytic hierarchy process based personalized
recommendations for tourist attractions over the internet. Expert Syst Appl 36:933–943
54. Huang Y, Bian L (2009) A Bayesian network and analytic hierarchy process based personalized
recommendations for tourist attractions over the internet. Expert Syst Appl 36:933–943
55. Huang Y, Bian L (2015) Using Ontologies and formal concept analysis to integrate heterogeneous
tourism information. IEEE Trans Emerging Topics Comput, 3, 2015
56. Ickjai L, Guochen C, Kyungmi L (2013) Points-of-interest mining from people’s photo-taking behavior.
In: Proceedings of the Annual Hawaii International Conference on System Sciences, 46:3129–1110
57. Jamali M, Ester M (2009) Using a trust network to improve top-n recommendation. In: Proceedings of
the third ACM conference on Recommender systems (pp 181–188)
58. Jannach D, Zanker M, Jessenitschnig M (2010) Developing knowledge-based travel advisor systems:
A case study. In: Sharda N (ed) Tourism Informatics: Visual Travel Recommender Systems, Social
Communities, and User Interface Design (pp 38–53). Hershey, PA: Information Science Reference
59. K-j K, Ahn H (2008) A recommender system using ga k-means clustering in an online shopping market.
Expert Syst Appl 34:1200–1209
60. Kapcak Ö, Spagnoli S, Robbemond V, Vadali S, Najafian S, Tintarev N (2018) Tourexplain: a
crowdsourcing pipeline for generating explanations for groups of tourists. In: Proceedings of the Work-
shop on Recommenders in Tourism, RecTour 2018, co-located with the 12th ACM Conference on
Recommender Systems (RecSys 2018), Vancouver, Canada, October, 7:33–36
61. Katsumi H, Yamada W, Ochiai K (2020) Generic poi recommendation. In: Adjunct Proceedings of the
2020 ACM International Joint Conference on Pervasive and Ubiquitous Computing and Proceedings
of the 2020 ACM International Symposium on Wearable Computers UbiComp-ISWC ’20 (p. 46–49).
New York, NY, USA: Association for Computing Machinery
62. Khatibi A, Belém F, da Silva APC, Almeida JM, Goncalves MA (2019) Fine-grained tourism
prediction: Impact of social and environmental features. Inform Process Manag, (p. 102057)
63. Kosmides P, Remoundou C, Demestichas K, Loumiotis I, Adamopoulou E, Theologou M (2014)
A location recommender system for location based social networks. In: International Conference on
Mathematics and Computers in Sciences and in Industry (pp 277–280)
64. Kotiloglu S, Lappas T, Pelechrinis K, Repoussis P (2017) Personalized multi-period tour recommen-
dations. Tour Manag 62:76–88
65. Kurashima T, Iwata T, Irie G, Fujimura K (2013) Travel route recommendation using geotagged photos.
Knowl Inf Syst 37:37–60
66. L C, C Y, W M (2009) Ontological recommendation multi-agent for tainan city travel. Expert Syst
Appl 36:6740–6753
67. Lamsfus C, Alzua-Sorzabal A, Martin D, Salvador Z, Usandizaga A (2009) Human-centric ontology-
based context modelling in tourism. In: Proceedings of the International Conference on Knowledge
Engineering and Ontology Development (pp 6–8). Madeira, Portugal, October, 2009, 424–434: Funchal
68. Lamsfus C, Alzua-Sorzabal A, Martin D, Smithers T (2011) An evaluation of a contextual approach to
visitor information system. In: Law R, Fuchs M, Ricci F (eds) Proceeding of the ENTER conference
(pp 191–202). Austria. January 2011, 179-189
69. Laß C., Herzog D, Wörndl W. (2017) Context-aware tourist trip recommendations. In: Proceed-
ings of the 2nd Workshop on Recommenders in Tourism co-located with 11th ACM Conference on
Recommender Systems (RecSys 2017), Como, Italy, August, 27:18–25
70. Lee H, Chung N, Nam Y (2019) Do online information sources really make tourists visit more diverse
places?: based on the social networking analysis. Inform Process Manag 56:1376–1390
9024 Multimedia Tools and Applications (2023) 82:8983–9027

71. Leng Y, Rudolph L, Pentland A, Zhao J, Koutsopoulos H (2016) Managing travel demand: location
recommendation for system efficiency based on mobile phone data. In: Proceedings of Data for Good
Exchange (D4GX) New York, NY. volume 2016
72. Li L, Lee KY, Yang S.-B. (2019) Exploring the effect of heuristic factors on the popularity of user-
curated ‘best places to visit’ recommendations in an online travel community. Inform Process Manag
56:1391–1408
73. Li X, Jiang M, Hong H, Liao L (2017) A time-aware personalized point-of-interest recommendation
via high-order tensor factorization. ACM Trans Inf. Syst. 35:31:1–31:23
74. Liao J, Tang J, Zhao X, Shang H (2018) Improving poi recommendation via dynamic tensor completion.
Sci Program 2018:11
75. Liliana A, Anna G, Giovanna P, Pietro MST (2003) Intrigue:, Personalized recommendation of tourist
attractions for desktop and handset devices. Appl Artif Intell 17:687–714
76. Lim KH (2016) Personalised recommendation of travel itineraries based on tourist interests and
preferences. In: Extended Proceedings of the 24th Conference on User Modeling
77. Lim KH (2016) Recommending and planning trip itineraries for individual travellers and groups of
tourists. In: Proceedings of the 26th International Conference on Automated Planning and Scheduling
(pp 115–120)
78. Lim KH, Chan J, Leckie C, Karunasekera S (2015) Personalized tour recommendation based on user
interests and points of interest visit durations. In: Proceedings of the Twenty-Fourth International Joint
Conference on Artificial Intelligence (pp 1778–1784)
79. Lim KH, Chan J, Leckie C, Karunasekera S (2016) Towards next generation touring: Personalized
group tours. In: Proceedings of the Twenty-Sixth International Conference on Automated Planning and
Scheduling (pp 412–420)
80. Linaza MT, Aguirregoikoa A, Garcia A, Torres JI, Aranburu K (2011) Image-based travel recommender
system for small tourist destinations. Information and Communication Technologies in Tourism, 1–12
81. Liu G.-Q., Zhang Y.-J., Fu Y.-M., Liu Y (2014) Behavior identification based on geotagged photo data
set. Sci World J 2014:616030
82. Liu Q, Ge Y, Li Z, Chen E, Xiong H (2011) Personalized travel package recommendation. In: 11th
IEEE International Conference on Data Mining (pp 407–416)
83. Loh S, Lorenzi F, Garin RS, Lichtnow D (2003) A tourism recommender system based on collaboration
and text analysis. Inform Technol Tourism 6:157–165
84. Lorenzi F, Abel M, Loh S, Aé P. (2011) Enhancing the quality of recommendations through expert
and trusted agents. In: 23rd IEEE International Conference on Tools with Artificial Intelligence
(pp 329–335)
85. Lorenzi F, Loh S, Abel M (2011) Personaltour: a recommender system for travel packages. In: Proceed-
ings of the 2011 IEEE/WIC/ACM International Conference on Intelligent Agent Technology. Lyon,
France, pp 22–27
86. Lu H-C, Lin C-Y, Tseng SV (2011) Trip-mine: An efficient trip planning approach with travel time
constraints. In: Proceedings - IEEE International Conference on Mobile Data Management (pp 152–
161)
87. Lu J, Wu D, Mao M, Wang W, Zhang G (2015) Recommender system application developments: a
survey. Decis Support Syst 74:12–32
88. Lu X, Wang C, J-m Y, Pang Y, Zhang L (2010) Photo2trip: generating travel routes from geo-tagged
photos for trip planning. ACM Multimed 2010:25–29
89. Lucas J, Luz N, Moreno M, Anacleto R, Figueiredo A, Martins C (2013) A hybrid recommendation
approach for a tourism system. Expert Syst Appl 40:3532–3550
90. Majid A, Chen L, Mirza HT, Hussain I, Chen G (2015) A system for mining interesting tourist locations
and travel sequences from public geo-tagged photos. Data Knowl Eng 95:66–86
91. Martinez-Santiago F, Ariza-Lopez F, Montejo-Raez A, Urena-Lopez A (2012) Geoasis: a knowledge-
based geo-referenced tourist assistant. Expert Syst Appl 39:11737–11745
92. Martinez-Santiago F, Ariza-Lpez F, Montejo-Raez A, Urena-Lopez A (2012) Geoasis: a knowledge-
based geo-referenced tourist assistant. Expert Syst Appl 39:11737–11745
93. Martinez-Torres MR, Toral SL (2019) A machine learning approach for the identification of the decep-
tive reviews in the hospitality sector using unique attributes and sentiment orientation. Tour Manag
75:393–403
94. Maruyama A, Shibata N, Murata Y, Yasumoto K, Ito M (2004) A personal tourism navigation system
to support traveling multiple destinations with time restrictions. In: 18th International Conference on
Advanced Information Networking and Applications
95. Meehan K, Lunney T, Curran K, McCaughey A (2013) Context-aware intelligent recommendation
system for tourism. Work Progress Session PerCom 19:328–331
Multimedia Tools and Applications (2023) 82:8983–9027 9025

96. Meehan K, Lunney T, Curran K, McCaughey A (2013) Context-aware intelligent recommendation

system for tourism. In: Proceedings of the 11th IEEE International Conference on Pervasive Computing
and Communications (pp 328–331)
97. Melville P, Sindhwani V (2010) Recommender systems. In: Encyclopedia of Machine Learning
(pp 829–838). Boston, MA: Springer US
98. Mezni H, Fayala M (2018) Time-aware service recommendation: taxonomy, review, and challenges.
Softw Pract Exp 48:2080–2108
99. MinFeng D, Athanasopoulos G (2011) Modelling australian domestic and international inbound travel:
a spatial-temporal approach. Tour Manag 32:1075–1084
100. Mobasher JX, B, Y Z (2004) Semantically enhanced collaborative filtering on the web. Web Mining:
From Web to Semantic Web, 57–76
101. Montejo-Raez A, Perea-Ortega JM, Garcia-Cumbreras MA, Martinez-Santiago F (2011) Otium: a web
based planner for tourism and leisure. Expert Syst Appl 38:10085–10093
102. Montejo-Raez A, Perea-Ortega JM, Garcia-Cumbreras MA, Martinez-Santiago F (2011) Otium: a web
based planner for tourism and leisure. Expert Syst Appl 38:10085–10093
103. Moreno A, Valls A, Isern D, Marin L, Borras J (2013) Sigtur/e-destination: ontology-based personalized
recommendation of tourism and leisure activities. Eng Appl Artif Intell 26:633–651
104. Mottini A, Lheritier A, Acuna-Agost R, Zuluaga MA (2018) Understanding customer choices to improve
recommendations in the air travel industry. In: Workshop on Recommenders in Tourism, pp, (pp 28–32)
105. Najafabadi MK, Mohamed A, Onn CW (2019) An impact of time and item influencer in collaborative
filtering recommendations using graph-based model. Inform Process Manag 56:526–540
106. O’Sullivan D, Smyth B, Wilson D (2003) Preserving recommender accuracy and diversity in sparse
datasets. Int J Artif Intell Tool 13:219–235
107. Palumbo E, Rizzo G, Troncy R, Baralis E (2017) Predicting your next stop-over from location-based
social network data with recurrent neural networks. In: Proceedings of the ACM RecSys Workshop on
Recommenders in Tourism ser. RecSys
108. Pan W, Chen S, Feng Z (2013) Automatic clustering of social tag using community detection. Appl
Math 7:675–681
109. Pazzani M (1999) A framework for collaborative, content-based and demographic filtering. Artif Intell
Rev 13:393–408
110. Pazzani M, Billsus D (2007) Content-based recommendation systems. Lect Notes Comput Sci, 4321
111. Pearl J (1988) Probabilistic reasoning in intelligent systems. Networks of plausible inference. Morgan
Kaufmann
112. Peng P, gu X, Zhu S, Shou L, Chen G (2019) One net to rule them all: efficient recognition and retrieval
of poi from geo-tagged photos. Multimedia Tools and Applications, 78
113. Rashid A, Farooq S, Abid A, Umer T, Bashir A, Zikria Y (2021) Social media intention mining for
sustainable information systems: categories, taxonomy, datasets and challenges. Complex & Intelligent
Systems, pp 1–27
114. Rawat KMS, YS (2017) Clicksmart: a context-aware viewpoint recommendation system for mobile
photography. IEEE Trans Circ Syst Video Technol 27:149–158
115. Rendle S, Freudenthaler C, Schmidt-Thieme L (2010) Factorizing personalized markov chains for
next-basket recommendation. In: Proceedings of 19th International Conference on World Wide Web
(pp 811–820). NC, USA, April 2010: Raleigh
116. Ricci F, Nguyen QN, Averjanova O (2010) Exploiting a map-based interface in conversational
recommender systems for mobile travelers. In: Sharda N (ed) Tourism Informatics: Visual Travel
Recommender Systems, Social Communities, and User Interface Design (pp 73–93). Hershey, PA:
Information Science Reference
117. Ruı́z-Martı́nez J. M., Minarro JA, Castellanos D, Garcia F, Valencia R (2011) Ontology population: an
application for the e-tourism domain. Int J Innov Comput Inf Control 7:1–19
118. Ruotsalo T, Haav K, Stoyanov A, Rochee S, Fanid E, Deliaic R, Makela E, Kauppinen T, Hyvonen E
(2013) Smartmuseum: A mobile recommender system for the web of data. Web Semantics: Sci Service
Agents World Wide Web 20:50–67
119. Sarkar JL, Majumder A (2021) A new point-of-interest approach based on multi-itinerary recommen-
dation engine. Expert Syst Appl, (p. 115026)
120. Sarkar JL, Majumder A, Panigrahi CR, Roy S (2020) Multitour: a multiple itinerary tourists recom-
mendation engine. Electron Commer Res Appl, 40, 100943
121. Savir A, Brafman R, Shani G (2013) Recommending improved configurations for complex objects
with an application in travel planning. In: Proceedings of the 7th ACM conference on Recommender
systems (pp 391–394)
122. web of science (2020) https://clarivate.com/webofsciencegroup/solutions/web-of-science: last
accessed:20/10/2020
9026 Multimedia Tools and Applications (2023) 82:8983–9027

123. Sebastia L, Garcia I, Onaindia E, Guzman C (2009) e-tourism: a tourist recommendation and planning
application. Int J Artif Intell Tool 18:717–738
124. Sebastia L, Garcia I, Onaindia E, Guzman C (2009) e-tourism: a tourist recommendation and planning
application. Int J Artif Intell Tool 18:717–738
125. Sebastia L, Giret A, Garcia I (2010) A multi agent architecture for tourism recommendation. Trends in
Practical Applications of Agents and Multiagent Systems, 547–554
126. Sebastia L, Giret A, Garcia I (2010) A multi agent architecture for tourism recommendation. trends in
practical applications of agents and multiagent systems. Advan Intell Soft Comput 71:547–554
127. Seidel I, Gartner M, P”ottler M, Berger H, Dittenbach M, Merkl W (2010) Itchy feet: a 3d e-tourism
environment. Tourism Informatics: Visual Travel Recommender Systems, Social Communities, and
User Interface Design, (pp 209–242)
128. Shambour Q, Lu J (2011) A hybrid trust-enhanced collaborative filtering recommendation approach
for personalized government-to-business e-services. Int J Intell Syst 26:814–843
129. Sieh HP, C-T L, S-D L (2015) Triprouter: a time-sensitive route recommender system. In: IEEE
International Conference on Data Mining Workshop (pp 1207–1210)
130. Smyth B, Cotter P (2000) A personalised tv listings service for the digital tv age. Knowl-Based Syst
13:53–59
131. SongJie G, HongWu Y, HengSong T (2009) Combining memory-based and model-based collaborative
filtering in recommender system. In: Proceedings of the 2009 Pacific-Asia Conference on Circuits (pp
690–693). PACCS 2009: Communications and System
132. Souffriau W, Vansteenwegen P, Vertommen J, Berghe GV (2008) A personalized tourist trip design
algorithm for mobile tourist guides. Appl Artif Intell, 964–985
133. Swagatam D, Abraham A, Konar A (2008) Automatic clustering using an improved differential
evolution algorithm. IEEE Trans Syst Man Cybern Part A: Syst Humans 38:218–237
134. T-H B, S-B P (2016) Point of interest mining with proper semantic annotation. Multimed Tool Appl
76:22
135. Tai C-H, Yang D-N, Lin L-T, Chen M-S (2008) Recommending personalized scenic itinerarywith geo-
tagged photos. In: 2008 IEEE International Conference on Multimedia and Expo (pp 1209–1212)
136. Tan Y, Zhu Y (2010) Fireworks algorithm for optimization. Int Conf Swarm Intell, 355–364
137. Uddin MR, Ravishankar CVJ (2011) Finding regions of interest from trajectory data. In: IEEE 12th
International Conference on Mobile Data Management (pp 39–48)
138. Umanets A, Ferreira A, Leite N (2014) Guideme - a tourist guide with a recommender system and
social interaction. Proc Technol 17:407–414
139. Valls A (2003) Clusdm: a multiple criteria decision method for heterogeneous data sets. PhD thesis
16:129–130
140. Vansteenwegen ODV, P (2007) The mobile tourist guide: an or opportunity. OR Insight, 21–27
141. Vansteenwegen SWBGV, V ODP (2011) The city trip planner: an expert system for tourists. Expert
Syst Appl, 6540–6546
142. Vicient C, Sanchez D, Moreno A (2013) An automatic approach for ontology-based feature extraction
from heterogeneous textual resources. Eng Appl Artif Intell 26:1093–1106
143. Wang W, Zeng G, Tang D (2011) Bayesian intelligent semantic mashup for tourism. Concurrency
Comput: Pract Exp 23:850–862
144. Wilson DC, Smyth B, O’Sullivan D (2003) Sparsity reduction in collaborative recommendation: a
case-based approach. Int J Pattern Recognit Artif Intell 17:863–884
145. Wu B, Murata Y, Shibata N, Yasumoto K, Ito M (2009) A method for composing tour schedules
adaptive to weather change. IEEE Intell Vehicles Symp, 1407–1412
146. Xiang S, Fu Y, Chen H, Ran W, Liu T (2020) Multi-level feature learning with attention for person
re-identification. Multimed Tool Appl 79:1–15
147. Xu-yin W, Xiang-pei H, Wei-guo L (2006) An urban tourism intelligent recommendation system based
on webgis. In: International Conference on Management Science and Engineering
148. Yager R (1988) On ordered weighted averaging operators. Oper Multi-criteria Decision Making
18:183–190
149. Yang WS, Hwang SY (2013) itravel: a recommender system in mobile peer-to-peer environment. J Syst
Softw 86:12–20
150. Yang X, Steck H, Guo Y, Liu Y (2012) On top-k recommendation using social networks. In:
Proceedings of the sixth ACM conference on Recommender systems (pp 67–74)
151. Yin H, Zhou X, Cui B, Wang H, Zheng K, Nguyen Q (2016) Adapting to user interest drift for poi
recommendation. IEEE Trans Knowl Data Eng 28:2566–2581
152. Yin Z, Cao L, Han J, Luo J, Huang T (2011) Diversified trajectory pattern ranking in geo-tagged social
media. In: Proceedings of the 2011 SIAM International Conference on Data Mining
Multimedia Tools and Applications (2023) 82:8983–9027 9027

153. Ying H, Wu J, Xu G, Liu Y, Liang T, Zhang X, Xiong H (2018) Time-aware metric embedding with
asymmetric projection for successive poi recommendation. World Wide Web, 1–16
154. Yoon H, Zheng Y, Xie X, Woo W (2010) Smart itinerary recommendation based on user-generated
gps trajectories. In: Proceedings of the 7th International Conference on Ubiquitous Intelligence and
Computing (pp 19–34)
155. Yu F, Cui L, Guo W, Lu X, Li Q, Lu H (2020) A category-aware deep model for successive poi
recommendation on sparse check-in data. In: Proceedings of The Web Conference 2020 WWW ’20
(p. 1264–1274). New York, NY, USA: Association for Computing Machinery
156. Yu Y, Chen X (2015) A survey of point-of-interest recommendation in location-based social networks.
Trajectory-Based Behavior Analytics: Papers from the, 2015, 53–60
157. Yu Y, Wang Y, Gao S, Tang Z (2017) Statistical modeling and prediction for tourism economy using
dendritic neural network. Comput Intell Neurosci 2017:1–10
158. Yu Z, Xu H, Yang Z, Guo B (2015) Personalized travel package with multi-point-of-interest recom-
mendation based on crowdsourced user footprints. IEEE Trans Human-Mach Syst 46:1–8
159. Yu Z, Xu H, Yang Z, Guo B (2016) Personalized travel package with multi-point-of-interest recom-
mendation based on crowdsourced user footprints. IEEE Trans Human-Mach Syst 46:151–158
160. Yuan F, Jose JM, Guo G, Chen L, Yu H, Alkhawaldeh RS (2016) Joint geo-spatial preference and pair-
wise ranking for point-of-interest recommendation. In: IEEE 28th International Conference on Tools
with Artificial Intelligence (ICTAI (pp 46–53)
161. Yuan J, Zheng Y, Xie X (2012) Discovering regions of different functions in a city using human
mobility and pois. In: Proceedings of the 18th ACM SIGKDD International Conference on Knowledge
Discovery and Data Mining (pp 186–194)
162. Zhang YC, Medo M, Ren J, Zhou T, Li T, Yang F (2007) Recommendation model based on opinion
diffusion. EPL (Europhysics Letters) 80:6
163. Zhao K, Zhang Y, Yin H, Wang J, Zheng K, Zhou X, Xing C (2020) Discovering subsequence patterns
for next poi recommendation. In: Bessiere C. (ed) Proceedings of the Twenty-Ninth International Joint
Conference on Artificial Intelligence, IJCAI-20 (pp 3216–3222). International Joint Conferences on
Artificial Intelligence Organization. Main track
164. Zhao S, Zhao T, Yang H, Lyu MR, King I (2016) Stellar: spatial-temporal latent ranking for succes-
sive point-of- interest recommendation. In: Proceedings of Thirtieth AAAI Conference on Artificial
Intelligence (pp 315–322). AZ, USA: Phoenix
165. Zheng Y, Xie X (2011) Learning travel recommendations from user-generated gps traces. ACM Trans
Intell. Syst. Technol. 2:2:1–2:29
166. Zhou W, Han W (2019) Personalized recommendation via user preference matching. Inform Process
Manag 56:955–968

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps
and institutional affiliations.

Affiliations

Joy Lal Sarkar1 · Abhishek Majumder1 · Chhabi Rani Panigrahi2 · Sudipta Roy3 ·
Bibudhendu Pati2

Joy Lal Sarkar

joylalsarkar@gmail.com
Chhabi Rani Panigrahi
panigrahichhabi@gmail.com
Sudipta Roy
sudipta.it@gmail.com
Bibudhendu Pati
patibibudhendu@gmail.com
1 Tripura University, Tripura, India
2 Rama Devi Women’s University, Bhubaneswar, India
3 Assam University, Silchar, India