CN112632404B - Next point of interest recommendation method based on multi-granularity self-attention - Google Patents
Next point of interest recommendation method based on multi-granularity self-attention Download PDFInfo
- Publication number
- CN112632404B CN112632404B CN202011616569.2A CN202011616569A CN112632404B CN 112632404 B CN112632404 B CN 112632404B CN 202011616569 A CN202011616569 A CN 202011616569A CN 112632404 B CN112632404 B CN 112632404B
- Authority
- CN
- China
- Prior art keywords
- poi
- sequence
- self
- attention
- decoder
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 43
- 230000008569 process Effects 0.000 claims abstract description 26
- 238000012549 training Methods 0.000 claims abstract description 25
- 230000008447 perception Effects 0.000 claims description 36
- 230000006870 function Effects 0.000 claims description 28
- 238000012545 processing Methods 0.000 claims description 14
- 230000002776 aggregation Effects 0.000 claims description 13
- 238000004220 aggregation Methods 0.000 claims description 13
- 238000012360 testing method Methods 0.000 claims description 13
- 230000004913 activation Effects 0.000 claims description 5
- 230000002194 synthesizing effect Effects 0.000 claims description 4
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 230000002123 temporal effect Effects 0.000 description 13
- 235000019580 granularity Nutrition 0.000 description 8
- 238000012546 transfer Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000006855 networking Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9536—Search customisation based on social or collaborative filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/01—Social networking
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Databases & Information Systems (AREA)
- Business, Economics & Management (AREA)
- Probability & Statistics with Applications (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
本发明公开了一种基于多粒度自注意力的下一个兴趣点推荐方法，其包括：构建多粒度自注意力网络模型，该网络模型包括：对时间进行感知编码的第一编码器、对空间进行感知编码的第二编码器、对POI特征进行解码的解码器；采集多个用户历史签到数据，对上述网络模型进行训练，利用训练好的网络模型预测用户的下一个兴趣点；训练和预测的过程包括：根据用户历史签到数据提取包含时间特征和空间特征的POI序列，根据用户历史签到数据提取包含时间特征和空间特征的POI商圈序列；将POI序列和POI商圈序列作为输入序列，通过上述网络模型处理得到用户下一个POI id。本发明在综合考虑POI序列和POI商圈序列后再应用多粒度自注意力网络，可以更准确的推荐下一个兴趣点。
The present invention discloses a method for recommending the next point of interest based on multi-granularity self-attention, which includes: constructing a multi-granularity self-attention network model, the network model includes: a first encoder for perceptual encoding of time, a second encoder for perceptual encoding of space, and a decoder for decoding POI features; collecting multiple user historical check-in data, training the above network model, and using the trained network model to predict the user's next point of interest; the training and prediction process includes: extracting a POI sequence containing time features and space features according to the user's historical check-in data, extracting a POI business district sequence containing time features and space features according to the user's historical check-in data; using the POI sequence and the POI business district sequence as input sequences, and obtaining the user's next POI id through the above network model. The present invention comprehensively considers the POI sequence and the POI business district sequence and then applies the multi-granularity self-attention network, so that the next point of interest can be recommended more accurately.
Description
Technical Field
The invention relates to the technical field of information processing. More particularly, the present invention relates to a method for next point of interest recommendation based on multi-granularity self-attention.
Background
With the rapid development of intelligent devices, social networks such as WeChat, microblog and Facebook attract billions of users to exchange and share information. In recent years, a significant advancement in social networking applications has been the introduction of location-based systems, which have led to the advent of location-based social networks (LBSNs), such as Fourdeure and Yelp. Point of interest (POI) recommendation systems play a vital role in social networks, which can help users find locally interesting attractions or facilities. In addition, it may help the service provider recommend personalized services (e.g., advertisements) based on the user's check-in sequence. However, most of the next point of interest recommendations only focus on the transfer pattern on the POI sequence, and the transfer pattern between POI business circles is ignored, resulting in low accuracy of the next point of interest recommendation.
Disclosure of Invention
It is an object of the present invention to solve at least the above problems and to provide at least the advantages to be described later.
Still another object of the present invention is to provide a method for recommending a next point of interest based on multi-granularity self-attention, which can more accurately recommend a next point of interest by applying a multi-granularity self-attention network after comprehensively considering a POI sequence and a POI business turn sequence.
To achieve these objects and other advantages and in accordance with the purpose of the invention, a method for recommending next points of interest based on multi-granularity self-attention is provided, comprising:
constructing a multi-granularity self-attention network model, wherein the multi-granularity self-attention network model comprises a first encoder for perceptually encoding time, a second encoder for perceptually encoding space and a decoder for decoding POI characteristics;
collecting historical check-in data of a plurality of users, training the multi-granularity self-attention network model by taking the historical check-in data of part of the users as a training set, taking the historical check-in data of the other part of the users as a test set, and predicting the next interest point of the users by utilizing the trained multi-granularity self-attention network model;
The training set training process and the testing set testing process comprise the steps of extracting POI sequences containing time features and space features according to user historical sign-in data, and extracting POI business circle sequences containing time features and space features according to user historical sign-in data;
and using the POI sequence and the POI business circle sequence as input sequences, and processing the POI sequence and the POI business circle sequence through the multi-granularity self-attention network model to obtain the next POI id of the user.
Preferably, the user history check-in data includes a user id, a POI id, a check-in time, a category to which the POI belongs, a POI category, and a POI geographical location.
Preferably, the decoder includes a first decoder decoding the category of the POI, a second decoder decoding the category to which the POI belongs, and a third decoder decoding the POI id.
Preferably, the first encoder and the second encoder are each comprised of a feature layer, an aggregation layer, and a self-attention network.
Preferably, the processing of the POI sequence and the POI business circle sequence through the multi-granularity self-attention network model to obtain the next POI id of the user comprises the following steps of:
using a first encoder to encode the time features in the POI sequence to obtain time perception encoding of the POI sequence, using a second encoder to encode the space features in the POI sequence to obtain space perception encoding of the POI sequence, and synthesizing the time perception encoding of the POI sequence and the space perception encoding of the POI sequence to obtain POI sequence representation;
The method comprises the steps of using a first encoder to encode time features in a POI business turn sequence to obtain time perception encoding of the POI business turn sequence, using a second encoder to encode space features in the POI business turn sequence to obtain space perception encoding of the POI business turn sequence, and integrating the time perception encoding of the POI business turn sequence and the space perception encoding of the POI business turn sequence to obtain POI business turn sequence representation;
Using the time perception code of the POI sequence and the time perception code of the POI business circle sequence as input quantities, and processing the input quantities by a first decoder to obtain a predicted value of the next POI type;
The aggregate representation of the time features in the POI sequence obtained in the encoding process and the aggregate representation of the time features in the POI business circle sequence obtained in the encoding process are used as input quantities, and the next POI belonging category predicted value is obtained through processing of a second decoder;
and taking the POI sequence representation, the POI business circle sequence representation, the predicted value of the next POI category and the predicted value of the category to which the next POI belongs as input quantities, and processing the input quantities by a third decoder to obtain the next POI id of the user.
Preferably, the self-attention network is added with a layer of feedforward neural network, and the activation function used is ReLu functions.
Preferably, the activation function of the first decoder is a sigmoid function, and the loss function is a cross entropy loss function.
Preferably, the loss function of the second decoder is a BPR loss function.
Preferably, the loss function of the third decoder is a BPR loss function.
Preferably, adaGrad optimizers are used in the multi-granularity self-attention network model training process.
The invention has the advantages that the POI sequence is used as a fine granularity sequence, the POI business circle sequence is used as a coarse granularity sequence, modeling is carried out on two layers of the coarse granularity sequence and the fine granularity sequence, the model expression capability can be effectively improved, a multi-granularity self-attention network is adopted to capture ordered transfer modes of two granularities, an activity task for predicting the category of the next interest point and an auxiliary task for predicting the category of the next interest point are introduced, and finally, the next interest point is recommended by integrating the results of the coarse granularity sequence representation and the two tasks.
Additional advantages, objects, and features of the invention will be set forth in part in the description which follows and in part will become apparent to those having ordinary skill in the art upon examination of the following or may be learned from practice of the invention.
Drawings
FIG. 1 is a flow chart of a next point of interest recommendation method based on multi-granularity self-attention according to the present invention;
FIG. 2 is a flow chart of the multi-granularity self-attention network model training and testing process of the present invention;
FIG. 3 is a flow chart of the multi-granularity self-attention network model encoding and decoding process of the present invention;
Fig. 4 is a diagram illustrating a history of access records of Alice of a user according to one embodiment of the invention.
Detailed Description
The present invention is described in further detail below with reference to the drawings to enable those skilled in the art to practice the invention by referring to the description.
It should be noted that the described embodiments of the present application are only some embodiments, but not all embodiments of the present application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
As shown in fig. 1, the present invention provides a method for recommending next interest points based on multi-granularity self-attention, which includes:
S101, constructing a multi-granularity self-attention network model, wherein the multi-granularity self-attention network model comprises a first encoder for perceptually encoding time, a second encoder for perceptually encoding space and a decoder for decoding POI characteristics;
Here the first encoder and the second encoder are each comprised of a feature layer, an aggregation layer and a self-attention network.
The self-attention network uses an attention mechanism to calculate the correlation between each element and all other elements. In order to calculate the self-attention score, there are mainly the following processes:
Firstly, obtaining Q, K and V according to the dot multiplication of an input matrix X and a weight matrix:
Q=XWQ;
K=XWK;
V=XWV;
Secondly, for Q, K and V, calculating, and normalizing the calculated result by a softmax function:
Wherein, The function of (2) is to scale the result. Since the previous calculations are all linear calculations, in order to make the model have better nonlinear fitting capability, and thus add a layer of feedforward neural network, the activation function used is ReLu function:
F=FFN(Z)=ReLu(ZW1+b1);
Finally, the result F of the self-attention network is obtained.
Here, the POI feature includes a POI id, a category to which the POI belongs, a POI category, and a POI geographical position, and the decoder that decodes the POI feature includes a first decoder that decodes the POI category, a second decoder that decodes the POI category, and a third decoder that decodes the POI id.
S102, collecting historical check-in data of a plurality of users, training the multi-granularity self-attention network model by taking the historical check-in data of part of the users as a training set, taking the historical check-in data of the other part of the users as a test set, and predicting the next interest point of the users by utilizing the trained multi-granularity self-attention network model;
For example, there are 10 users, each having some history, and assuming p 1～p10 for training, in the training set, p 1～p9 is used as training to predict p 10 for each user, and assuming that the ratio of the training set to the testing set is 8:2, that is, 8 user histories are used for training, to obtain the learnable parameters of the multi-granularity self-attention network model, and 2 user histories are used as testing set for performing the test alone to predict the next interest point of the user. Of course, this is merely illustrative, and in the practical application process, the number of training sets is continuously increased along with the increase of the number of users, so that the prediction result of the multi-granularity self-attention network model is more and more accurate.
The training set training process and the testing set testing process herein include:
S201, extracting POI sequences containing time features and space features according to user historical sign-in data, and extracting POI business circle sequences containing time features and space features according to user historical sign-in data;
Here, the user history check-in data includes a user id (u), a POI id (p), a check-in time (t), a category (c) to which the POI belongs, a POI category (y), and a POI geographical location (g), and the usage parameters are expressed as (u, p, t, c, y, g).
The user id (u), the POI id (p), the check-in time (t) and the POI geographic position (g) are better understood, and the user id (u), the POI id (p), the check-in time (t) and the POI geographic position (g) can be known according to the literal meaning, and the category (c) to which the POI belongs can be roughly classified into catering, hospitals, education and the like. For POI category (y), we explain with the following example:
As shown in fig. 4, the user Alice has the following history access record p 2→p3→…→p8 on the electronic map. Wherein the individual POIs p 2,p3,p4,p5 are among the business turn POIs p 1. In this model, the POI sequence (fine-grained sequence) is p 2→p3→…→p8 and the POI business turn sequence (coarse-grained sequence) is p 1→p6→p7→p8, where p 2→p3→p4→p5 is reduced to business turn POI p 1. Since individual POI p 6,p7,p8 has no corresponding POI business at a higher granularity, we will still keep p 6,p7,p8 in order to preserve the integrity of the sequence context semantics. If the POI category (y) is distinguished by 0 and 1, the value of the POI p 2～p8 category (y) is 0 in the POI sequence and the value of the category (y) to which p 1 belongs is 1 in the POI business sequence.
In the sequence of the POI,For the temporal characteristics of user u at time n,Is a spatial feature of user u at time n. Similarly, in the POI business turn sequence,For the temporal characteristics of user u at time n,Is a spatial feature of user u at time n.
S202, using the POI sequence and the POI business circle sequence as input sequences, and processing through the multi-granularity self-attention network model to obtain the next POI id of the user.
The steps specifically comprise the following steps:
s301, encoding the time features in the POI sequence by using a first encoder to obtain the time perception encoding of the POI sequence.
In the sequence of the POI,For the temporal feature of user u at time n, the first encoder consists of a feature layer, an aggregation layer, and a self-attention network.
First, the temporal feature aggregation is expressed as:
Wherein, Is of the kind embedding (of the genus embedding),Is a embedding time stamp map to 24 hours,Is embedding of the category to which it belongs, B 1 is a learnable parameter.
The aggregated representation of the temporal feature is then fed into a self-attention network, resulting in a time-aware encoding of the POI sequence:
Wherein SAN represents a self-care network. Represented is a time-aware encoding of the POI sequence.
And using a second encoder to encode the spatial features in the POI sequence to obtain spatial perception encoding of the POI sequence, and synthesizing the temporal perception encoding of the POI sequence and the spatial perception encoding of the POI sequence to obtain POI sequence representation.
In the sequence of the POI,For the spatial feature of user u at time n, it is noted that it is not directly used in the process of extracting the featureI.e. the spatial coordinates, but the distance between the current position and the last position is recorded asThe second encoder is also composed of a feature layer, an aggregation layer, and a self-attention network.
First, the spatial feature aggregation is expressed as:
Wherein, For the POI id to correspond to embedding,For the interval of the current position from the last position, B 2 is a learnable parameter.
Next, the process willThe type of (2) is set to be an integer type, so that the computational complexity is reduced, and then the aggregate representation of the spatial features is sent into a self-attention network to obtain the spatial perception coding of the POI sequence:
Wherein SAN represents a self-care network. Represented is a spatial perceptual coding of a POI sequence. Finally, combining the time perception coding and the space perception coding of the POI sequence into a POI sequence representation:
S302, encoding the time features in the POI business turn sequence by using a first encoder to obtain the time perception encoding of the POI business turn sequence
In the sequence of the POI business turn,Is the temporal characteristic of user u at time n. The first encoder is comprised of a feature layer, an aggregation layer, and a self-attention network.
First, the temporal feature aggregation is expressed as:
Wherein, Is of the kind embedding (of the genus embedding),Is a embedding time stamp map to 24 hours,Is embedding of the category to which it belongs, B 3 is a learnable parameter.
Then, the aggregate representation of the temporal features is fed into a self-attention network, resulting in a temporal perceptual coding of the POI business turn sequence:
wherein the SAN represents a self-care network, Represented is a time-aware encoding of the POI business turn sequence.
Using a second encoder to encode the spatial features in the POI business turn sequence to obtain spatial perception encoding of the POI business turn sequence, and synthesizing the temporal perception encoding of the POI business turn sequence and the spatial perception encoding of the POI business turn sequence to obtain POI business turn sequence representation;
in the sequence of the POI business turn, Is a spatial feature of user u at time n. Notably, they are not directly used in the process of extracting featuresI.e. the spatial coordinates, but the distance between the current position and the last position is recorded asThe second encoder is also composed of a feature layer, an aggregation layer, and a self-attention network.
First, the spatial feature aggregation is expressed as:
Wherein, For the POI id to correspond to embedding,For the interval of the current position from the last position, B 4 is a learnable parameter.
Next, the process willThe type of (2) is set as an integer, so that the computational complexity is reduced, and then the aggregate representation of the spatial features is sent into a self-attention network to obtain the spatial perception coding of the POI business turn sequence:
Wherein SAN represents a self-care network. Representing the spatial perceptual coding of a sequence of POI business circles. Finally, combining the time perception coding and the space perception coding of the POI business turn sequence to form a POI business turn sequence representation:
S303, using the time perception code of the POI sequence and the time perception code of the POI business circle sequence as input quantity, and processing the input quantity by a first decoder to obtain a predicted value of the next POI type;
to predict the next POI category, we use temporal perceptual coding from two sequences to represent:
Wherein, Is an aggregation of the temporal perceptual encoders by two sequences, σ is a sigmoid function,Is a predicted value for the next POI category.
For the prediction of the next POI category we use Cross-entropy (Cross-entropy) loss function:
s304, using an aggregate representation obtained by the time features in the POI sequence in the encoding process and an aggregate representation obtained by the time features in the POI business circle sequence in the encoding process as input quantities, and obtaining a predicted value of the category to which the next POI belongs through processing of a second decoder;
In order to predict the category to which the next POI belongs, we first calculate the time-coded aggregate representation from the two sequences, then calculate the aggregate a rep of the two time-coded aggregate representations, and finally calculate the category to which the next POI belongs:
wherein A rep is a time-aware representation, For the representation of the user's hand,And the predicted value of the category to which the next POI belongs.
For the prediction of the category to which the next POI belongs, we use the BPR loss function:
S305, using the POI sequence representation, the POI business circle sequence representation, the predicted value of the next POI category and the predicted value of the category to which the next POI belongs as input quantities, and obtaining the next POI id of the user through processing of a third decoder.
In order to recommend the next POI, we aggregate the POI sequence representation, the POI business circle sequence representation, the predicted value of the category of the next POI and the predicted value of the category of the next POI, then calculate the aggregate P rep of the two sequence space coding aggregate representations, and finally recommend the next POI id:
Wherein, Is an aggregation of the first encoder result and the second encoder result of different sequences,Is a multi-granularity sequence representation, P rep is a spatially aware representation,For the representation of the user's hand,Predicted values for the next POI id of the user.
For recommending the next POI, we use the BPR loss function:
For the total loss function, there is the following representation:
Where λ y,λc and λ p are used to adjust for different losses, λ y+λc+λp =1, λ is the regularization coefficient, Θ= (W, b) is the set of learnable parameters.
Here, when steps S303 to 305 are performed in the training process, a AdaGrad optimizer is further used to optimize the training process of the learnable parameters.
According to the method in the embodiment, the POI sequence is used as the fine-granularity sequence, the POI business circle sequence is used as the coarse-granularity sequence, modeling is carried out on two levels of the coarse-granularity sequence and the fine-granularity sequence, the model expression capability can be effectively improved, a multi-granularity self-attention network is adopted to capture ordered transfer modes of two granularities, an activity task for predicting the category of the next interest point and an auxiliary task for predicting the category of the next interest point are introduced, and finally, the result of the coarse-granularity sequence representation and the two tasks is integrated to recommend the next interest point.
Although embodiments of the present invention have been disclosed above, it is not limited to the details and embodiments shown and described, it is well suited to various fields of use for which the invention would be readily apparent to those skilled in the art, and accordingly, the invention is not limited to the specific details and illustrations shown and described herein, without departing from the general concepts defined in the claims and their equivalents.
Claims (6)
1. A method for recommending next points of interest based on multi-granularity self-attention, comprising:
constructing a multi-granularity self-attention network model, wherein the multi-granularity self-attention network model comprises a first encoder for perceptually encoding time, a second encoder for perceptually encoding space and a decoder for decoding POI characteristics;
collecting historical check-in data of a plurality of users, training the multi-granularity self-attention network model by taking the historical check-in data of part of the users as a training set, taking the historical check-in data of the other part of the users as a test set, and predicting the next interest point of the users by utilizing the trained multi-granularity self-attention network model;
The training set training process and the testing set testing process comprise the steps of extracting POI sequences containing time features and space features according to user historical sign-in data, and extracting POI business circle sequences containing time features and space features according to user historical sign-in data;
Using the POI sequence and the POI business circle sequence as input sequences, and processing the POI sequence and the POI business circle sequence through the multi-granularity self-attention network model to obtain the next POI id of the user;
The user history sign-in data comprises user id, POI id, sign-in time, category of the POI, POI category and POI geographic position;
the decoder comprises a first decoder for decoding the category of the POI, a second decoder for decoding the category of the POI and a third decoder for decoding the POI id;
the first encoder and the second encoder are composed of a feature layer, an aggregation layer and a self-attention network;
The process of obtaining the next POI id of the user through the multi-granularity self-attention network model processing by taking the POI sequence and the POI business circle sequence as input sequences comprises the following steps:
using a first encoder to encode the time features in the POI sequence to obtain time perception encoding of the POI sequence, using a second encoder to encode the space features in the POI sequence to obtain space perception encoding of the POI sequence, and synthesizing the time perception encoding of the POI sequence and the space perception encoding of the POI sequence to obtain POI sequence representation;
The method comprises the steps of using a first encoder to encode time features in a POI business turn sequence to obtain time perception encoding of the POI business turn sequence, using a second encoder to encode space features in the POI business turn sequence to obtain space perception encoding of the POI business turn sequence, and integrating the time perception encoding of the POI business turn sequence and the space perception encoding of the POI business turn sequence to obtain POI business turn sequence representation;
Using the time perception code of the POI sequence and the time perception code of the POI business circle sequence as input quantities, and processing the input quantities by a first decoder to obtain a predicted value of the next POI type;
The aggregate representation of the time features in the POI sequence obtained in the encoding process and the aggregate representation of the time features in the POI business circle sequence obtained in the encoding process are used as input quantities, and the next POI belonging category predicted value is obtained through processing of a second decoder;
and taking the POI sequence representation, the POI business circle sequence representation, the predicted value of the next POI category and the predicted value of the category to which the next POI belongs as input quantities, and processing the input quantities by a third decoder to obtain the next POI id of the user.
2. The multi-granularity self-attention based next point of interest recommendation method as recited in claim 1, wherein said self-attention network incorporates a layer of feed-forward neural network using an activation function of ReLu.
3. The multi-granularity self-attention based next point of interest recommendation method of claim 1, wherein the activation function of the first decoder is a sigmoid function and the loss function is a cross entropy loss function.
4. The multi-granularity self-attention based next point of interest recommendation method of claim 1, wherein the loss function of the second decoder is a BPR loss function.
5. The multi-granularity self-attention based next point of interest recommendation method of claim 1, wherein the loss function of the third decoder is a BPR loss function.
6. The multi-granularity self-attention based next point of interest recommendation method as recited in claim 1, wherein a AdaGrad optimizer is used in the multi-granularity self-attention network model training process.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011616569.2A CN112632404B (en) | 2020-12-30 | 2020-12-30 | Next point of interest recommendation method based on multi-granularity self-attention |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011616569.2A CN112632404B (en) | 2020-12-30 | 2020-12-30 | Next point of interest recommendation method based on multi-granularity self-attention |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112632404A CN112632404A (en) | 2021-04-09 |
| CN112632404B true CN112632404B (en) | 2024-12-27 |
Family
ID=75287137
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011616569.2A Active CN112632404B (en) | 2020-12-30 | 2020-12-30 | Next point of interest recommendation method based on multi-granularity self-attention |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112632404B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113205427B (en) * | 2021-06-07 | 2022-09-16 | 广西师范大学 | Recommendation method for next interest point of social network |
| CN114201683B (en) * | 2021-12-15 | 2025-06-27 | 齐鲁工业大学(山东省科学院) | A method and system for recommending news based on interest activation based on multi-level matching |
| CN115982477B (en) * | 2022-12-05 | 2025-06-10 | 北京电子工程总体研究所 | Method for recommending next interest point based on space-time information representation |
| CN115827974B (en) * | 2022-12-05 | 2025-06-27 | 北京理工大学 | Next interest point recommendation system based on space-time information representation |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111931067A (en) * | 2020-09-14 | 2020-11-13 | 北京百度网讯科技有限公司 | Interest point recommendation method, device, equipment and medium |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108875007B (en) * | 2018-06-15 | 2019-12-17 | 腾讯科技（深圳）有限公司 | method and device for determining interest point, storage medium and electronic device |
-
2020
- 2020-12-30 CN CN202011616569.2A patent/CN112632404B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111931067A (en) * | 2020-09-14 | 2020-11-13 | 北京百度网讯科技有限公司 | Interest point recommendation method, device, equipment and medium |
Non-Patent Citations (1)
| Title |
|---|
| Context-aware Attention-based Data Augmentation for POI Recommendation;Yang Li(et al.);2019 IEEE 35th International Conference on Data Engineering Workshops (ICDEW);20190701;第177-184页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112632404A (en) | 2021-04-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112632404B (en) | Next point of interest recommendation method based on multi-granularity self-attention | |
| Jiao et al. | A novel next new point-of-interest recommendation system based on simulated user travel decision-making process | |
| CN111400620A (en) | User Trajectory Position Prediction Method Based on Spatio-temporal Embedding Self-Attention | |
| CN109062962B (en) | Weather information fused gated cyclic neural network interest point recommendation method | |
| CN110399565A (en) | Points of interest recommendation method based on recurrent neural network based on spatio-temporal periodic attention mechanism | |
| Luca et al. | Deep learning for human mobility: a survey on data and models | |
| CN108804646B (en) | Point of interest sign-in prediction method integrating deep learning and factorization machine | |
| CN105825297A (en) | Markov-model-based position prediction method | |
| CN113139140A (en) | Tourist attraction recommendation method based on space-time perception GRU and combined with user relation preference | |
| CN106776928A (en) | Recommend method in position based on internal memory Computational frame, fusion social environment and space-time data | |
| CN115827974B (en) | Next interest point recommendation system based on space-time information representation | |
| CN114862010B (en) | A method, device, equipment and medium for determining flow based on spatiotemporal data | |
| Shi et al. | RegNet: a neural network model for predicting regional desirability with VGI data | |
| CN106909560B (en) | Interest point ordering method | |
| CN117633371B (en) | Recommendation method, device and readable storage medium based on multi-attention mechanism | |
| CN112650949B (en) | Regional POI (point of interest) demand identification method based on multi-source feature fusion collaborative filtering | |
| Moon et al. | A Large-Scale Study in Predictability of Daily Activities and Places. | |
| CN118379138A (en) | Risk assessment method, risk assessment device, computer equipment and computer readable storage medium | |
| CN117609631A (en) | A fine-grained point of interest recommendation method and device based on a few-sample knowledge graph | |
| CN116991908A (en) | Interest point recommendation method based on multivariable Hawkes space-time point process embedded attention | |
| CN115774819A (en) | A Point of Interest Recommendation Method and System Based on Hierarchical Recurrent Neural Network | |
| Chen et al. | A multi-context aware human mobility prediction model based on motif-preserving travel preference learning | |
| CN115827898A (en) | User intention-oriented spatio-temporal knowledge graph construction method | |
| Guo et al. | Fine-grained dynamic price prediction in ride-on-demand services: Models and evaluations | |
| CN115482019A (en) | Activity attention prediction method and device, electronic equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |