CN114969566B - Distance-measuring government affair service item collaborative filtering recommendation method - Google Patents

Abstract

A distance measurement government affair service affair collaborative filtering recommendation method comprises the steps of respectively constructing a government affair service affair portrait, a user basic portrait and a user behavior portrait; calculating the similarity between users by adopting the portrait; calculating the relative position between the user and the government affair service center according to the user position information expression mode; and calculating the quasi-recommendation score of the target user for the related matters according to the user similarity and the distance between the user and the matters, thereby recommending the related matters. The user transaction frequency calculation method provided by the invention can fully reflect the correlation degree of the user and the transaction service items, introduces the spatial neighborhood relation and the distance calculation method, and participates in item score calculation; the method can preferentially recommend items of the same place for the user, improve the accuracy of the recommendation result, make up the problem of insufficient consideration of the space position by the traditional collaborative filtering, and develop and enrich the collaborative filtering recommendation method.

Description

A collaborative filtering recommendation method for government service items based on distance measurement

technical field

The invention belongs to the technical field of information recommendation, and specifically relates to a collaborative filtering recommendation method for government service items based on distance measurement, which combines the characteristics of government service items, spatial distance and collaborative filtering methods to recommend government service items for natural and legal person users.

Background technique

In recent years, the State Council has attached great importance to the work of "Internet + government services", and has issued a series of policy documents to promote the standardization, standardization and facilitation of government services. On the one hand, the state organizes government departments at all levels to comprehensively sort out administrative power matters and public service matters, compile and form a list of government service matters, standardize the guidelines for handling government service matters, and lay a foundation for enterprises and the public to handle affairs conveniently. On the other hand, various regions have strengthened the construction of government service networks, and promoted more government service matters to be handled online, on the palm, and at fingertips. At present, more and more government service matters have been handled online, and the public can directly handle matters through the Internet, which simplifies the handling process, saves handling time, greatly improves the efficiency of handling affairs, and effectively enhances the sense of gain of the corporate masses.

When business people handle government service matters through the Internet, they need to find the matters to be handled on the webpage first. Due to the large number of catalogs of online government service items and professional classification, it is difficult for ordinary users to quickly and accurately locate the items they need. Although some websites already provide search functions, they mainly use keyword matching to locate queries, without comprehensive consideration of user characteristics and historical behavior, resulting in inaccurate recommendation results. In fact, in the commercial or academic field, recommendation systems have been developed to solve the problem of information overload, such as shopping websites, video websites, and academic retrieval. Users can quickly locate the desired content according to the results recommended by the system.

Collaborative filtering is a commonly used algorithm in current recommendation methods. It assumes that two users A and B have similar behavior habits (such as purchasing, reading, watching movies, etc.), then they also have similar preferences on other items. The collaborative filtering algorithm does not need to know user preferences, and only uses user history behavior to predict the user's rating of unknown products for recommendation. The method is simple and effective, and has been applied in recommender systems in many fields.

In the recommendation of government service items, the use of collaborative filtering algorithm is a good choice, but there are still some problems. First, the government service processing business scenario is different from commercial shopping. User ratings are not closely related to user services. Blindly applying the rating matrix to calculate user similarity will affect the recommendation accuracy. Second, government services have significant spatial attributes. Generally, government service items require users to be approved by the competent department of the place of household registration or residence. When using a collaborative filtering algorithm, if the location information is used as a user label, valid information may be filtered out, and if the location information is not considered, the recommendation result may be inaccurate.

Therefore, how to apply the collaborative filtering method to the recommendation of government service items, avoid filtering useful information, and improve the accuracy of recommendation has become a technical problem that needs to be solved urgently.

Contents of the invention

Aiming at the needs of online government service handling, the present invention combines the collaborative filtering recommendation method with the characteristics of government service items, and fully considers user location information, and proposes a distance-measured government service item collaborative filtering recommendation method and device. For reaching this purpose, the present invention adopts following technical scheme:

For reaching this purpose, the present invention adopts following technical scheme:

A collaborative filtering recommendation method for government service items based on distance measurement, characterized in that it includes the following steps:

Data preparation step S110:

Based on government service item data, user registration data, and user handling behavior data, respectively construct government service item portraits, user basic portraits, and user behavior portraits;

User similarity calculation step S120:

Calculate the similarity 𝑠𝑖𝑚 (𝑘, 𝑘 ′ ) between users 𝑘 and 𝑘' by using the user's basic portrait, user's work behavior portrait and government service item portrait;

Distance calculation step S130:

According to the expression method of user location information, choose administrative division or European distance to calculate the distance d𝑖𝑠 _𝑘𝑗 between user 𝑘 and the government service center handling matters 𝑗;

Recommended item score calculation step S140:

According to the user similarity 𝑠𝑖𝑚 (𝑘, 𝑘 ′ ), the distance d𝑖𝑠 _𝑘𝑗 between the user 𝑘 and the handling item 𝑗 , calculate the proposed recommendation score of the target user 𝑘 for related items, the formula is as follows:

(7)

(7)

Among them, 𝑠𝑐𝑜𝑟𝑒 _𝑘𝑗 represents the score of the government service items 𝑗 that the user 𝑘' has handled for the target user 𝑘, where 𝑤 represents the weight;

Result recommendation step S150:

According to the score result of the items to be recommended in S140, the top N items are selected and recommended to the user in order of the scores from high to low.

Optionally, in the data preparation step S110:

The portrait of the government service item is based on the attribute information of the government service item, tags the item, and builds a model. The attribute information of the government service item includes: item name, handling department, service object, handling location, and item theme;

The user basic portrait is to use the basic information filled in when the user registers to tag the user and build a model. The user registration data is specifically, the basic information of a natural person includes gender, age, occupation, marital status and location, legal person Basic information may include business scope, business type, industry, business size and business address;

The user behavior portrait is the user's service behavior data, and the user's service behavior is tagged to construct a model. The user's service behavior data includes the attribute information of the user's handling items and the number of times the user has handled them.

Optionally, in step S120,

The cosine similarity calculation method is used to calculate the similarity between users 𝑠𝑖𝑚 (𝑘, 𝑘 ′ ),

（1）

(1)

表示所有用户对标签𝑇特征 指数的平均值。 Among them, 𝐿 represents the collection of all user basic portrait tags, 𝑇 represents a specific label in the user basic portrait, 𝑀 _𝑘𝑇 and 𝑀 _𝑘'𝑇 respectively represent the feature index of user 𝑘 and 𝑘' on the label 𝑇,

Represents the average of all users' feature indices for the tag 𝑇.

Optionally, in step S120,

The user characteristic index is used to reflect the correlation between user tags and handling items, and the calculation formula is as follows:

(2)

(2)

In the formula, 𝑀 _𝑘𝑇 indicates the user characteristic index of user 𝑘 under the user basic portrait label 𝑇, 𝐼𝑡𝑒𝑚 _𝑘 indicates the items handled by user 𝑘, 𝑁(𝐼𝑡𝑒𝑚 ₎ indicates the total number of items handled by user 𝑘, 𝑟 _𝑘𝑘 indicates the number of items handled by user 𝑘 𝑗’s work frequency, 𝐼 _𝑗𝑇 represents the user’s basic portrait label 𝑇 index of the handling matter 𝑗.

Optionally, in step S120,

The user's work frequency is used to reflect the degree of association between the user and the handling matter, and is calculated according to the user's behavior portrait. The formula is as follows:

(3)

(3)

Among them, 𝑟 _𝑘𝑗 indicates the frequency with which user 𝑘 handles items 𝑗, 𝑁(𝑗 _𝑘 ) indicates the number of times users 𝑘 handles items 𝑗, and 𝑁(𝐼𝑡𝑒𝑚 _𝑘 ) indicates the total number of items handled by user 𝑘.

Optionally, in step S120,

The user basic portrait label index is used to judge the relevance between the user's label and the handling item, the formula is as follows:

(4)

(4)

Among them, 𝐼 _𝑗𝑇 represents the user base portrait tag 𝑇 index of the handling matter 𝑗, 𝑁(𝑈 _𝑗 ) is the number of all users handling the matter 𝑗, 𝑁(𝑇 _𝑗 ) represents the number of users who have the label 𝑇 among the users handling the matter 𝑗.

Optionally, in the distance calculation step S130,

When the user location information is represented by coordinates, the Euclidean distance is used to calculate the distance d𝑖𝑠 _𝑘𝑗 between the user 𝑘 and the government service center handling the matter 𝑗, the formula is as follows:

(5)

(5)

Among them, (𝑥 _𝑘 , 𝑦 _𝑘 ) indicates the coordinate position of the user 𝑘, (𝑥 _𝑗 , 𝑦 _𝑗 ) indicates the coordinate position of the government affairs service center of the processing item 𝑗;

When the user location information is represented by administrative divisions, the distance d𝑖𝑠 _𝑘𝑗 between the user 𝑘 and the government service center that handles the matter 𝑗 is calculated according to the spatial relationship of the administrative divisions, as follows:

(6)

(6)

Among them, (𝑥 _𝑎 , 𝑦 _𝑎 ) indicates the coordinate position of the center point of the administrative division where the user 𝑘 is located, and (𝑥 _𝑏 , 𝑦 _𝑏 ) indicates the coordinate position of the center point of the area where the government service center that handles the matter 𝑗 is located.

The present invention further discloses a storage medium, the storage medium is used for storing computer-executable instructions, and when the computer-executable instructions are executed by a processor, the above-mentioned collaborative filtering recommendation method for government service items based on distance measurement is executed.

The present invention has the following advantages:

First, when calculating user similarity, a calculation method of user service frequency (Formula 3) is proposed, which replaces the user rating in traditional collaborative filtering, and can fully reflect the degree of association between users and handling government service items, and is more suitable for government affairs Service item recommendation scenario.

Second, introduce the spatial neighborhood relationship and distance calculation method (Formula 5-7) in the recommendation process of government service items, define the distance between the user and the handling item according to point coordinates or administrative divisions, and participate in item scoring calculate.

The distance measure proposed by the present invention reflects the territorial requirements of the administrative department when dealing with matters, and can preferentially recommend matters of the same territoriality for users, thereby improving the accuracy of recommendation results. At the same time, the distance measure also makes up for the lack of consideration of the spatial location of the traditional collaborative filtering, and develops and enriches the collaborative filtering recommendation method.

Description of drawings

Fig. 1 is a flowchart of a collaborative filtering recommendation method for government service items based on distance measurement according to a specific embodiment of the present invention.

Detailed ways

The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention, but not to limit the present invention. In addition, it should be noted that, for the convenience of description, only some structures related to the present invention are shown in the drawings but not all structures.

The invention mainly lies in: adapting the collaborative filtering algorithm to the government service field. At the same time, the focus is on increasing the spatial distance measure. When recommending items, user similarity and location attributes are considered comprehensively to improve recommendation accuracy.

Specifically, referring to FIG. 1 , it shows a flow chart of a collaborative filtering recommendation method for government service items based on distance measurement according to the present invention, including the following steps:

Data preparation step S110:

Based on the government service item data, user registration data and user handling behavior data, the government service item portrait, user basic portrait, and user behavior portrait are respectively constructed.

Preferably, the portrait of the government service item is based on the attribute information of the government service item, and the item is tagged to construct a model. The government service matters mentioned include: the name of the matter, the handling department, the service object, the place of handling and the subject of the matter.

Preferably, the basic user profile is to use the basic information filled in by the user when registering to tag the user and build a model. The user registration data is specifically, the basic information of a natural person includes gender, age, occupation, marriage and childbearing status, location, etc., and the basic information of a legal person may include business scope, enterprise type, industry, enterprise scale, enterprise address, etc.

Preferably, the user behavior portrait is user service behavior data, and the user service behavior is tagged to construct a model. The user service behavior data includes attribute information of the user's handling items and the number of user handling times.

User similarity calculation step S120:

User similarity The user judges the similarity between the basic attributes and service needs of two users. Through the user similarity, similar users can be found for the target user, and it is convenient to recommend similar users to the target user.

Specifically, this step is:

The similarity between users 𝑘 and 𝑘' is calculated by using information such as user basic portraits, user handling behavior portraits, and government service item portraits.

Specifically, the cosine similarity calculation method is used to calculate the similarity 𝑠𝑖𝑚 (𝑘, 𝑘 ′ ) between users.

（1）

(1)

表示所有用户对标签𝑇特征 指数的平均值。𝑠𝑖𝑚 (𝑘, 𝑘′)的值越大，则代表用户𝑘与𝑘'越相近。 Among them, 𝐿 represents the collection of all user basic portrait tags, 𝑇 represents a specific label in the user basic portrait, 𝑀 _𝑘𝑇 and 𝑀 _𝑘'𝑇 respectively represent the feature index of user 𝑘 and 𝑘' on the label 𝑇,

Represents the average of all users' feature indices for the tag 𝑇. The larger the value of 𝑠𝑖𝑚 (𝑘, 𝑘 ′ ), the closer the user 𝑘 is to 𝑘'.

Specifically, the user characteristic index is used to reflect the correlation between user tags and handling items, and the calculation formula is as follows:

(2)

(2)

In the formula, 𝑀 _𝑘𝑇 represents the user characteristic index of user 𝑘 under the user basic portrait label 𝑇. 𝐼𝑡𝑒𝑚 _𝑘 indicates the items handled by the user 𝑘, 𝑁(𝐼𝑡𝑒𝑚 _𝑘 ) indicates the total number of items handled by the user 𝑘, 𝑟 _𝑘𝑗 indicates the frequency of the user 𝑘's handling of the item 𝑗, and 𝐼 _𝑗𝑇 index indicates the label of the user who handles the basic item 𝑗 The larger the value of 𝑀 _𝑘𝑇 , the higher the relevance of user label 𝑇 to user 𝑘's business handling.

The user's service frequency is used to reflect the degree of association between the user and the handling item, which can be calculated according to the user's behavior portrait, and the formula is as follows:

(3)

(3)

Among them, 𝑟 _𝑘𝑗 indicates the frequency with which user 𝑘 handles items 𝑗, 𝑁(𝑗 _𝑘 ) indicates the number of times users 𝑘 handles items 𝑗, and 𝑁(𝐼𝑡𝑒𝑚 _𝑘 ) indicates the total number of items handled by user 𝑘. The higher the 𝑟 _𝑘𝑗 , the higher the frequency of handling the government service item, and the closer the user's business relationship with the item.

Preferably, the user's basic portrait label index is used to determine the relevance between the user's label and the handling item, and the formula is as follows:

(4)

(4)

Among them, 𝐼 _𝑗𝑇 represents the user base portrait label 𝑇 index of the handling matter 𝑗, 𝑁(𝑈 _𝑗 ) is the number of all users handling the matter 𝑗, 𝑁(𝑇 _𝑗 ) represents the number of users who have the label 𝑇 among the users handling the matter 𝑗, 𝐼 The larger the value of _𝑗𝑇 , the stronger the business relationship between the handling item 𝑗 and the user portrait label 𝑇.

Distance calculation step S130:

There are obvious geographical requirements for users to handle government services. For example, some government services must be handled by the competent department of the user's household registration or the user's residence. Therefore, measuring the positional relationship between users and items is of great significance for accurately recommending items.

There are two ways to represent government service items and user locations. One is to use the name of the administrative division of the area, such as Haidian District and Fengtai District; the other is to use specific coordinates. Generally, the government service item can be obtained in both ways of representing the location. The distance calculation is based on the representation of the user's location information.

According to the expression method of user location information, the relative distance d𝑖𝑠 _𝑘𝑗 between the user 𝑘 and the government service center handling the matter 𝑗 is calculated by using the administrative division or the Euclidean distance.

When the user location information is expressed in coordinates, the relative distance between the user and the government service center is calculated using the Euclidean distance, and the formula is as follows:

(5)

(5)

Among them, d𝑖𝑠 _𝑘𝑗 indicates the distance between the user 𝑘 and the government service center handling the matter 𝑗, (𝑥 _𝑘 , 𝑦 _𝑘 ) indicates the coordinate position of the user 𝑘, (𝑥 _𝑗 , 𝑦 _𝑗 ) indicates the coordinates of the government service center handling the matter 𝑗 Location.

When the user location information is represented by administrative divisions, according to the relative distance between the user and the government service center according to the spatial relationship of the administrative divisions, the specific design is as follows:

(6)

(6)

The same area means that the administrative division where the user 𝑘 is located is the same as the administrative division where the government service center that handles the matter 𝑗 is located, for example, both are located in Haidian District; the adjacent area refers to the administrative division where the government service center that handles the matter 𝑗 is located Adjacent areas, for example, if the user is located in Haidian District, the administrative division for handling matters is Xicheng District, that is, the adjacent areas can be the upper, lower, left, right, upper left, lower left, upper right, and lower right adjacent to this area eight regions. 𝑘 and j have no adjacency relationship means that there is no adjacency relationship between two administrative divisions, such as Xicheng District and Shijingshan District.

Among them, d𝑖𝑠 _𝑘𝑗 indicates the distance between the user 𝑘 and the government service center that handles the item 𝑗, (𝑥 _𝑎 , 𝑦 _𝑎 ) indicates the coordinate position of the center point of the administrative division where the user 𝑘 is located, (𝑥 _𝑏 , 𝑦 _𝑏 ) indicates the handling item The coordinate position of the center point of the area where j's government affairs service center is located.

Recommended item score calculation step S140:

The main idea of recommending government service items in the present invention is to recommend to the target user the items handled by the user with the same preferences as the target user.

According to the user similarity 𝑠𝑖𝑚 (𝑘, 𝑘 ′ ), the distance d𝑖𝑠 _𝑘𝑗 between user 𝑘 and the government service center handling item j, calculate the proposed recommendation score of target user 𝑘 for related items, the formula is as follows:

(7)

(7)

Among them, 𝑠𝑐𝑜𝑟𝑒 _𝑘𝑗 represents the score of the government service item j that user 𝑘' has handled for the target user 𝑘, where 𝑤 represents the weight, 𝑠𝑖𝑚 (𝑘, 𝑘 ′ ) represents the similarity between user 𝑘 and user 𝑘', d𝑖𝑠 _𝑘 Indicates the distance between user 𝑘 and the government service center handling item j.

Result recommendation step S150:

According to the score result of the items to be recommended in S140, the top N items are selected and recommended to the user in order of the scores from high to low.

Further, the present invention also discloses a storage medium, the storage medium is used to store computer-executable instructions, and when the computer-executable instructions are executed by a processor, the above-mentioned collaborative filtering recommendation of government service items based on distance measurement is performed method.

The present invention has the following advantages:

First, when calculating user similarity, a calculation method of user service frequency (Formula 3) is proposed, which replaces the user rating in traditional collaborative filtering, and can fully reflect the degree of association between users and handling government service items, and is more suitable for government affairs Service item recommendation scenario.

Second, introduce the spatial neighborhood relationship and distance calculation method (Formula 5-7) in the recommendation process of government service items, define the distance between the user and the handling item according to point coordinates or administrative divisions, and participate in item scoring calculate.

The distance measure proposed by the present invention reflects the territorial requirements of the administrative department when dealing with matters, and can preferentially recommend matters of the same territoriality for users, thereby improving the accuracy of recommendation results. At the same time, the distance measure also makes up for the lack of consideration of the spatial location of the traditional collaborative filtering, and develops and enriches the collaborative filtering recommendation method.

Obviously, those skilled in the art should understand that each unit or each step of the present invention described above can be realized by a general-purpose computing device, they can be concentrated on a single computing device, and optionally, they can be implemented by a program executable by the computer device codes, so that they can be stored in a storage device and executed by a computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps can be made into a single integrated circuit module for realization. As such, the present invention is not limited to any specific combination of hardware and software.

The above content is a further detailed description of the present invention in conjunction with specific preferred embodiments. It cannot be determined that the specific embodiments of the present invention are limited thereto. Under the present invention, several simple deduction or substitutions can also be made, all of which should be considered as belonging to the protection scope of the present invention determined by the submitted claims.

Claims (3)

1. A distance-measuring government affair service item collaborative filtering recommendation method is characterized by comprising the following steps:

data preparation step S110:

respectively constructing a government affair service affair portrait, a user basic portrait and a user behavior portrait based on government affair service affair data, user registration data and user affair handling behavior data;

user similarity calculation step S120:

calculating the similarity sim (k, k ') between the users k and k' by adopting the user basic portrait, the user transaction portrait and the government affair service affair portrait;

distance calculation step S130:

according to the useThe user position information expression mode selects administrative division or European distance to calculate the distance dis between the user k and the government affair service center handling the affair j _kj ；

Recommended item score calculation step S140:

according to the similarity sim (k, k') of the user and the distance dis between the user k and the transaction item j _kj And calculating the quasi-recommendation score of the target user k for the related matters, wherein the formula is as follows:

wherein, score _kj Represents a score for the government service event j that user k' has transacted for target user k, where w represents a weight;

result recommending step S150:

according to the item-to-be-recommended score result of S140, selecting the top N items to recommend to the user according to the order of the scores from high to low:

in the data preparation step S110:

the government affair service affair portrait is to label affairs according to attribute information of government affair service affairs, and construct a model, wherein the attribute information of government affair service affairs includes: item name, transaction department, service object, transaction location, and item topic;

the basic portrait of the user is that the basic information filled in during user registration is utilized to perform labeling processing on the user and construct a model, the user registration data is specifically that the basic information of natural people comprises gender, age, occupation, marriage and education conditions and positions, and the basic information of legal people can comprise an operation range, an enterprise type, an affiliated industry, an enterprise scale and an enterprise address;

the user behavior portrait is user transaction behavior data, labeling is carried out on the user transaction behavior, a model is constructed, and the user transaction behavior data comprises attribute information of user transaction items and user transaction times;

in the step S120, the process proceeds,

calculating the similarity sim (k, k') between users by using a cosine similarity calculation method,

where L represents the set of all user base representation tags, T represents a specific tag in the user base representation, M _kT And M _k′T Respectively representing the characteristic indexes of users k and k' to tag T,

representing the average value of all users to the tag T characteristic indexes;

the user characteristic index is used for reflecting the relevance between the user label and the transaction item, and the calculation formula is as follows:

in the formula, M _kT Item representing user characteristic index of user k under user basic portrait tag T _k Item N (Item) representing a transaction by user k _k ) Indicates the total number of transactions, r, transacted by user k _kj Indicating the transaction frequency, I, of the user k to the transaction item j _jT A user basic portrait label T index representing transaction item j;

the user transaction frequency is used for reflecting the degree of association between the user and transaction items, and is calculated according to the user behavior portrait, and the formula is as follows:

wherein r is _kj Indicates the transaction frequency of the user k to the item j, N (j) _k ) Indicates the number of transactions, N (Item), for Item j by user k _k ) Representing the total number of items transacted by the user k;

the user basic portrait label index is used for judging the relevance between the label of the user and the transaction items, and the formula is as follows:

wherein, I _jT User base portrait label T index, N (U) representing transaction item j _j ) Number of all users handling item j, N (T) _j ) Indicating the number of users having the tag T among the users transacting the item j.

2. The recommendation method according to claim 1,

in the distance-calculating step S130,

when the user position information is expressed by coordinates, the distance dis between the user k and the government affair service center handling the item j is calculated by adopting the Euclidean distance _kj The formula is as follows:

wherein (x) _k ，y _k ) Represents the coordinate position of user k, (x) _j ，y _j ) The coordinate position of the government affair service center for representing the transaction item j;

when the user position information is represented by administrative divisions, the distance dis between the user k and the government affair service center handling the item j is calculated according to the spatial relationship of the administrative divisions _kj The method comprises the following steps:

wherein (x) _a ，y _a ) Coordinate position (x) representing the center point of the administrative division in which user k is located _b ，y _b ) And (3) the coordinate position of the center point of the area where the government affairs service center for handling the affair j is positioned.

3. A storage medium, characterized by:

the storage medium storing computer-executable instructions which, when executed by a processor, perform the distance metric, government services event collaborative filtering recommendation method of claim 1 or 2.

Images