An information recommendation method for a government affairs service platform

Abstract

The present invention relates to an information recommendation method for a government affairs service platform, comprising the following steps: constructing a relational representation vector of the user linking to the page according to the historical records of the user accessing the pages of the government affairs service platform; mapping the relational representation vector to the tangent plane of the hyperbolic space to obtain hyperbolic features; using the hyperbolic features as neurons of a BP neural network to train the BP neural network to obtain an information recommendation model for recommending required pages for the user based on the user's historical records. The purpose of the present invention is to be more comprehensive when learning the behavioral characteristics of the user's history, so as to make the subsequent prediction and recommendation more accurate.

Description

Technical Field

[0001] The present invention relates to the technical field of big data processing, and particularly relates to an information recommendation method for a government service platform. Background Art

[0002] With the development of information technology, government service work is realizing electronic informatization as much as possible, improving the processing efficiency of government service work for users and government staff, and also having the traceability of processing, greatly improving the use efficiency of users for government services.

[0003] However, the existing government service pages often focus on providing standardized service options, ignoring the personalized needs and behavior habits of users, resulting in lack of pertinence and accuracy in service push. Nowadays, machine learning is an important means to realize big data processing. By analyzing the historical behavior of users (such as clicks, browsing, etc.), it can accurately predict user preferences and provide customized content, significantly improving user satisfaction. However, when machine learning extracts user behavior features, it is not accurate enough, resulting in a decrease in the accuracy of subsequent prediction recommendations. Summary of the Invention

[0004] The purpose of the present invention is to be more comprehensive when learning the historical behavior characteristics of users, so that subsequent prediction recommendations can be more accurate, and to provide an information recommendation method for a government service platform.

[0005] In order to achieve the above-mentioned invention purpose, the embodiments of the present invention provide the following technical solutions:

[0006] An information recommendation method for a government service platform, including the following steps:

[0007] Step 1, according to the historical records of users accessing the government service platform pages, construct a relationship representation vector of users linked to the pages;

[0008] Step 2, map the relationship representation vector to the tangent plane of the hyperbolic space to obtain hyperbolic features;

[0009] Step 3, use the hyperbolic features as the neurons of the BP neural network, train the BP neural network to obtain an information recommendation model for recommending the required pages for users based on the historical records of users.

[0010] The specific steps of Step 1 include the following steps:

[0011] Step 1-1, use U to represent the user set U = {u1,...u m ,u M}, where u m represents the m-th user, and M represents the number of users; user Z represents the page set of the government service platform Z = {z1,...z n ,...,zN}, where z n represents the nth page, and N represents the number of pages; if user u m has visited page z n , then construct the link matrix of user u m visiting page z n . ;

[0012] Step 1-2: Based on the link matrix , calculate the relationship characterization index S m of user u n with respect to page z mn ;

[0013] Step 1-3: Process the relationship characterization index S mn through the Embedding layer and convert it into a relationship characterization vector .

[0014] The specific steps of Step 1-1 are as follows:

[0015] If user u m has visited page z n , then the link matrix , construct the link matrix of user u m visiting page z n : :

[0016]

[0017] where is the number of interactions; is the number of sessions; is the number of entries; is the duration of the most recent stay; is the longest stay duration; is the reference duration; is the page type feature vector; is the page score; is the entry frequency.

[0018] The specific steps of Step 1-2 are as follows:

[0019] The first relationship characterization index :

[0020]

[0021] where v = 1, 2, 3; represents the browsing duration coefficient, used to balance the magnitude of ;

[0022] The second relationship representation index :

[0023]

[0024] wherein represents the type feature vector of page z n , r represents the type of page z n ;

[0025] The third relationship representation index :

[0026]

[0027] The relationship representation index S of user u m with respect to page z n is: mn

[0028]

[0029] wherein are respectively weight coefficients, used to balance the importance degrees of the first relationship representation index, the second relationship representation index, and the third relationship representation index.

[0030] Step 2 specifically includes the following steps:

[0031] Let H d represent a d-dimensional hyperbolic space with curvature c = -1 and constant k = -1 / c, and set the reference point as ; Map the relationship representation vector to the tangent plane of the hyperbolic space:

[0032]

[0033] wherein represents the hyperbolic feature of the relationship representation vector ; ||.|| L represents the L-order norm; [u m , z n represents the access operation of user u m linking to page z n ; cosh represents the hyperbolic cosine function; sinh represents the hyperbolic sine function.

[0034] Step 2 further includes the following steps:

[0035] Calculate the hyperbolic features for the pages accessed by M users, thereby obtaining the hyperbolic feature matrix :

[0036] ​

[0037] Hyperbolic feature matrix The m-th row in it represents user u m For the hyperbolic features of all pages, m = 1, 2,..., M; if user u m has not visited some pages, the corresponding hyperbolic features are 0.

[0038] Step 3 specifically includes the following steps:

[0039] Take the non-zero hyperbolic features as the neurons in the input layer of the BP neural network. If there are L non-zero hyperbolic features, the number of neurons in the input layer is L; through the non-linear transformation of the hidden layer, map the input data {x1, x2,..., x L} to a higher-dimensional feature space, where x l represents the l-th input data, l = 1, 2,..., L; through backpropagation, the weights of the hidden layer neurons can be gradually adjusted to minimize the overall loss of the network and finally complete the learning task;

[0040] Construct the loss function Loss of the BP neural network:

[0041]

[0042] where, x l represents the l-th input data in the input layer, l = 1, 2,..., L; y n represents the probability of the page z n recommended for the user. If x l represents the hyperbolic feature of the m-th user for the n-th page, then y n represents the probability of recommending the n-th page for the m-th user; represents the weight coefficient, ; ||.||2 represents the second norm; represents the Hadamard product.

[0043] Compared with the prior art, the beneficial effects of the present invention:

[0044] Through the historical access records of users to the government service platform, the present invention obtains the behavioral data of the pages accessed by users, constructs a link matrix, and calculates the relationship representation index of users to the pages based on the elements in the link matrix. This index is calculated through three indicators and can fully represent the behavioral characteristics of users. After converting the relationship representation index into a vector form, it is mapped to the tangent plane of the hyperbolic space. By utilizing the geometric characteristics of the hyperbolic space, the data of the hierarchical structure or tree-like relationship between users and each page can be represented more efficiently. Finally, the hyperbolic features output by the hyperbolic space are used as the neurons of the BP neural network to train the BP neural network. During the training process, the constructed loss function enables the network to fit the complex data distribution of the pages. The present invention combines the hyperbolic features mapped by the hyperbolic space with the neurons of the BP neural network, enabling machine learning to fully learn the behavioral characteristics of users towards web pages and the hierarchical structure between these web pages. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the embodiments. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.

[0046] Figure 1 It is a flowchart of the method of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some of the embodiments of the present invention, not all of them. Usually, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the detailed description of the embodiments of the present invention provided in the following drawings is not intended to limit the scope of the present invention to be protected, but only represents the selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0048] It should be noted that similar reference numerals and letters denote similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of the present invention, the terms "first", "second", etc. are only used for distinguishing descriptions and cannot be construed as indicating or implying relative importance or suggesting any actual relationship or order between these entities or operations. Additionally, the terms "connected" and "coupled" etc. can be directly connected between components or indirectly connected via other components.

[0049] Embodiment 1:

[0050] The present invention is implemented through the following technical solutions. As Figure 1 shown, an information recommendation method for a government service platform includes the following steps:

[0051] Step 1, construct a relational representation vector of the user linked to the page according to the historical record of the user accessing the government service platform page.

[0052] Use U to represent the user set U = {u1,...u m ,u M}, where u m represents the m-th user and M represents the number of users. User Z represents the page set Z = {z1,...z n ,...,z N}, where z n represents the n-th page and N represents the number of pages. The pages of the government service platform include various types, such as advertisements, news, notifications, filling in forms, etc. Filling in forms means filling in the personal information of the user through the government service platform to achieve government service processing. Users can query some real-time events through the local government service platform, such as traffic operation conditions, weather conditions, housing sales conditions, etc.

[0053] If user u m has accessed page z n , then the link matrix , construct the link matrix m of user u n accessing page z :

[0054]

[0055] wherein, is the number of interactions, indicating the number of times user u m clicks on any link in page z n ;

[0056] is the number of sessions, indicating user um The number of messages posted on page z n ;

[0057] is the number of entries, indicating that user u m enters page z n times;

[0058] is the stay duration, indicating that user u m last stayed on page z n (in seconds);

[0059] is the longest stay duration, indicating that user u m stayed on page z n for the longest single time (in seconds);

[0060] is the reference duration, indicating the reference duration (in seconds) for browsing page z n ;

[0061] is the page type feature vector, and page types include advertisements, news, notifications, forms (if the form information requires multiple pages, multiple pages for one item are regarded as one page), etc.;

[0062] is the page score, indicating the score given by user u m to page z n (there are four score levels: excellent 0.8, good 0.6, medium 0.4, poor 0.2);

[0063] is the entry frequency, indicating the frequency that user u m enters page z n ;

[0064] Based on the link matrix , calculate the relationship characterization index of user u m to page z n :

[0065] The first relationship characterization index :

[0066]

[0067] where v = 1, 2, 3; represents the browsing duration coefficient, used to balance the magnitude of .

[0068] The second relationship characterization index :

[0069]

[0070] Among them, represents the type feature vector of page z n and r represents the type of page z n .

[0071] The third relationship characterization index :

[0072]

[0073] For user u m the relationship characterization index S of page z n is: mn For:

[0074]

[0075] Among them, are respectively weight coefficients, used to balance the importance of the first relationship characterization index, the second relationship characterization index, and the third relationship characterization index.

[0076] After processing the relationship characterization index S mn through the Embedding layer, it is converted into a relationship characterization vector . Through the calculation of the above various indexes, the relationship characterization vector can fully represent the various behavioral characteristics of user u m for page z n . And this scheme only calculates the pages visited by user u m . For example, among N pages, only a part of the pages are visited by user u m , then only the relationship characterization indexes corresponding to these pages are calculated, and for the unvisited distributed pages, the link matrix is expressed as .

[0077] Step 2, map the relationship characterization vector to the tangent plane of the hyperbolic space to obtain hyperbolic features.

[0078] Let H d represent a d-dimensional hyperbolic space with curvature c = -1 and constant k = -1 / c, and set the reference point as . Map the relationship characterization vector to the tangent plane of the hyperbolic space:

[0079]

[0080] Among them, represents the hyperbolic feature of the relationship characterization vector ; ||.||L denotes the L - norm; [u m ,z n represents the access operation of user u m linking to page z n ; cosh represents the hyperbolic cosine function; sinh represents the hyperbolic sine function.

[0081] Since there are M users and N pages, the hyperbolic features are calculated for each page accessed by each user, thus obtaining the hyperbolic feature matrix :

[0082]

[0083] The hyperbolic feature matrix The m - th row in it represents the hyperbolic features of user u m for all pages, where m = 1, 2,..., M. If user u m has not accessed some pages, the corresponding hyperbolic features are 0. For example, if user u m has not accessed page z n-1 , then .

[0084] This solution uses the technique of hyperbolic space mapping to embed the behavioral feature data of users for pages into a hyperbolic geometric space with negative curvature. By utilizing the geometric properties of the hyperbolic space, it can more efficiently represent the data of the hierarchical or tree - like relationships between the pages accessed by users.

[0085] Step 3: Use the hyperbolic features as the neurons of the BP neural network, train the BP neural network to obtain an information recommendation model for recommending the required pages for users based on their historical records.

[0086] The hyperbolic feature matrix already contains the hyperbolic features of each user for pages. Take the non - zero hyperbolic features as the neurons of the input layer of the BP neural network. If there are L non - zero hyperbolic features, the number of neurons in the input layer is L. Then, through the non - linear transformation of the hidden layer, map the input data {x1, x2,..., x L} to a higher - dimensional feature space (where x l represents the l - th input data, l = 1, 2,..., L), gradually extract deep features, adjust the weights and biases, so that the network can fit the complex data distribution. Through backpropagation, the weights of the hidden - layer neurons can be gradually adjusted to minimize the overall loss of the network and finally complete the learning task.

[0087] Construct the loss function Loss of the BP neural network:

[0088]

[0089] where x l represents the l-th input data of the input layer, l = 1, 2, ..., L; y n represents the probability of the page z n being recommended to the user. If x l represents the hyperbolic feature of the m-th user for the n-th page, then y n represents the probability of recommending the n-th page to the m-th user; represents the weight coefficient, ; ||.||2 represents the second norm; represents the Hadamard product.

[0090] This solution maps the records of the user's historical access to the pages of the government service platform in the hyperbolic space, learns the user's behavioral characteristics of the pages, including the browsing duration, the number of browsing times, and the type of the page, etc., maps the user's behavioral characteristics of the pages into high-dimensional features that can be learned by the machine, trains a BP neural network through a loss function, and obtains an information recommendation model, so that when the user logs in to the government service platform in the future, the required pages can be recommended to the user.

[0091] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. An information recommendation method for a government service platform, characterized in that: The following steps are involved: Step 1: Based on the user's history of visiting the government service platform page, a relationship representation vector linking the user to the page is constructed; The step 1 specifically comprises the following steps: Step 1-1, let U represent the user set U={u1,...u m ,u M }, where u m represents the mth user, M represents the number of users; user Z represents the page set Z={z1,...z n ,...,z N }, where z n represents the nth page, N represents the number of pages; if user u m Visited pages n , then construct user u m Visit page n The link matrix ; The step 1-1 specifically includes the following steps: If user u m Visited pages n , then the link matrix , construct user u m Visit page n The link matrix : in, is the number of interactions; is the number of sessions; is the number of entries; The duration of the most recent stay; is the maximum length of stay; For reference duration; is the page type feature vector; Rate the page; is the frequency of entry; Step 1-2, based on the link matrix , calculate user u m For page z n The relationship representation index S mn ; The steps 1-2 specifically include the following steps: The first relational indicator : Among them, v=1,2,3; Indicates browsing time coefficient, used for balancing The size of ; Second relationship characterization index : in, Indicates page z n The type feature vector, r represents page z n Type; The third relationship representation index : User m For page z n The relationship representation index S mn for: in, They are The weight coefficient is used to balance the importance of the first relationship representation index, the second relationship representation index and the third relationship representation index; Step 1-3: Relationship representation index S mn After being processed by the Embedding layer, it is converted into a relation representation vector ; Step 2, mapping the relationship representation vector to the tangent plane of the hyperbolic space to obtain the hyperbolic feature; Step 3: Use the hyperbolic features as neurons of the BP neural network, train the BP neural network, and obtain an information recommendation model for recommending required pages to the user based on the user's historical records.

2. The information recommendation method of the government service platform according to claim 1, characterized in that: The step 2 specifically includes the following steps: Let H d Represents a d-dimensional hyperbolic space with curvature c=-1 and constant k=-1 / c. Set the reference point to ; Represent the relationship vector Mapping to the tangent plane of hyperbolic space: in, Representing the relationship vector Hyperbolic characteristics of ;||.|| L represents the L-order norm; [u m ,z n ] indicates user u m Link to page z n access operation; cosh represents the hyperbolic cosine function; sinh represents the hyperbolic sine function.

Images