Bid invitation information recommendation method and system, electronic equipment and readable storage medium

Abstract

The invention discloses a bid invitation information recommendation method, electronic equipment and a readable storage medium. The method comprises the steps of collecting bid invitation item information and user information; generating an initial representation vector for each user and each bid invitation item through the embedded layer; inputting the initial representation vector into a multi-layer embedded propagation layer, and distributing weights for neighbor nodes through a multi-head attention mechanism to obtain an updated representation vector of the user and the bid invitation project node; inputting the plurality of updated representation features into a feature fusion layer, distributing different weights for the updated representation vector by the feature fusion layer, and performing weighted summation to obtain a final representation vector; inputting the final representation vector into a prediction layer, wherein the prediction layer obtains a prediction score between the user and the bid invitation item by using a two-layer multi-layer perceptron network; and selecting a plurality of bid invitation items with the highest prediction score to recommend to the user.

Description

Technical Field

[0001] The present invention relates to the technical field of content recommendation, and in particular to a bidding information recommendation method and system, an electronic device and a readable storage medium. Background Art

[0002] Tendering and bidding are an integral part of business activities and are of great significance in ensuring the fairness, efficiency and economic benefits of transactions. In this process, the tendering party publicly invites suppliers or contractors to submit their quotations for completing specific work or providing specific goods and services, and suppliers or contractors submit quotations and related conditions in accordance with the tendering requirements in order to obtain transaction opportunities.

[0003] With the development of economy and the improvement of information level, the number of bidding projects is increasing. Enterprises need to find bidding information that matches them from the massive information. In the fast-paced business environment, quickly obtaining bidding information is crucial to seize business opportunities. Therefore, enterprise users need an efficient way to obtain bidding information to meet the different needs and preferences of enterprise users. However, the existing recommendation methods have the problems of inaccurate recommendations and low efficiency.

[0004] In view of the above problems, designing an efficient and accurate bidding information recommendation method is an urgent problem to be solved by technical personnel in this field. Summary of the invention

[0005] In view of this, the embodiments of the present invention provide a bidding information recommendation method and system, an electronic device and a readable storage medium, which recommend bidding project information matching the user to improve the efficiency and accuracy of recommending bidding project information.

[0006] In order to achieve the purpose of the invention, the present invention adopts the following technical solutions: In a first aspect, a bidding information recommendation method is provided, characterized in that the method comprises: Collecting bidding project information and user information, wherein the bidding project information is obtained by crawling the bidding website; Generate initial representation vectors for each user and bidding item through the embedding layer; The initial representation vector is input into a multi-layer embedding propagation layer. The embedding propagation layers at different layers learn the structural information of neighbors with different hop counts. The embedding propagation layer assigns weights to neighbor nodes through a multi-head attention mechanism to aggregate neighbor node information and obtain updated representation vectors of user and bidding project nodes. The multi-layer embedding propagation layer outputs multiple updated representation vectors. Inputting the multiple updated representation features into a feature fusion layer, the feature fusion layer assigns different weights to the updated representation vectors of the embedding propagation layers at different layers, and performing weighted summation of the updated representation vectors of the embedding propagation layers at different layers to obtain a final representation vector; Inputting the final representation vector into a prediction layer, wherein the prediction layer uses a two-layer multilayer perceptron network to obtain a prediction score between the user and the bidding item; A number of bidding projects with the highest prediction scores are selected and recommended to the user.

[0007] Optionally, in the above-mentioned method embodiments of the present invention, the bidding project information is obtained by crawling from a bidding website, including: The bidding project information is collected using a bidding information collection system. After the bidding information collection system is configured, the bidding information collection system automatically crawls the latest data from the bidding website every day and extracts the characteristics of the bidding project.

[0008] Optionally, in the above method embodiments of the present invention, the configuration includes the starting URL of the website, the URL of the subsequent page turning, related request parameters, data extraction rules and time extraction rules.

[0009] Optionally, in the above method embodiments of the present invention, the bidding project characteristics include the bidding project number, publishing area, publishing unit, bidding unit, agency, publishing time, bidding type, bidding product, amount, and bidding deadline.

[0010] Optionally, in the above-mentioned method embodiments of the present invention, user characteristics are extracted from the user information, and the user characteristics include user basic information, user qualification information, user financial information, user legal proceedings information, and user historical bidding information.

[0011] Optionally, in the above-mentioned method embodiments of the present invention, the basic information of the user includes the business scope, registered capital and location of the user's company; The user qualification information includes user qualification level, credit level, and professional qualification personnel information, wherein the professional qualification personnel information includes the number, level, and years of experience of professional qualification personnel; The user's financial information includes financial reports in recent years; The user's legal litigation information includes the user's involvement in litigation and arbitration in recent years; The user's historical bidding information includes the user's bidding information and the winning project data.

[0012] Optionally, in the above method embodiments of the present invention, generating an initial representation vector for each user and bidding item through an embedding layer includes: Create user IDs and bidding project IDs for users and bidding projects respectively, and build an embedding matrix based on the user IDs and bidding project IDs; Encode user features and bidding project features to build user feature embedding vectors and the bidding item feature embedding vector , R d is a d-dimensional matrix consisting of user and bidding item feature embedding vectors, where l represents the lth feature; The user's initial representation vector is , n is the number of users, and the initial representation vector of the bidding project is , m is the number of bidding projects.

[0013] Optionally, in the above-mentioned method embodiments of the present invention, the embedded propagation layers of the different layers learn structural information of neighbors with different hop counts, including: The embedded propagation layer has an N-layer structure, and the j-th embedded propagation layer corresponds to a neighbor j hops away from the entity, where N is an integer greater than 1, j=1, 2, ..., N.

[0014] Optionally, in the above-mentioned method embodiments of the present invention, the embedding propagation layer assigns weights to neighbor nodes through a multi-head attention mechanism to achieve aggregation of neighbor node information, and obtains updated representation vectors of user and bidding project nodes, including: In the j-th layer of the embedding propagation layer, the set of j-hop neighbor nodes of a user u is ,in, l is the number of neighboring nodes of i hops, j=1,2,…,N, N is the number of layers of the embedded propagation layer, and N is an integer greater than 1; The weight of user u to neighbor node v1 is calculated through the attention mechanism: Among them, π v1 is the weight of user u to neighbor node v1, u T is the transpose of the vector matrix of user u, e u and e v1 are the initial representation vectors of user u and neighbor node v1 respectively, and W is the transformation matrix; Use the softmax function for normalization, expressed as: in, is the normalized weight, e vk Represents the initial representation vector of the neighbor node vk; The neighbor node set representation of user u is obtained by weighted addition of all neighbor nodes, which is expressed as: Among them, e Nu is the neighbor node set of user u, is the normalized weight, e vk Represents the initial representation vector of the neighbor node vk; Aggregate the initial representation vector of user u and the neighbor node set representation of user u, expressed as: Among them, agg is the aggregation vector, LeakyReLU() is the LeakyReLU activation function, W is the conversion matrix, b is the offset, e u is the initial representation vector of user u, e Nu is the neighbor node set of user u; Under the multi-head attention mechanism, the aggregate vectors obtained under each attention mechanism are concatenated to obtain the updated representation vector of the user of the j-th layer of the embedding propagation layer, which is expressed as: in, is the updated representation vector of the user in the j-th embedding propagation layer, P is the number of heads of the attention mechanism, ⊕ is the concatenation operation, and σ() is the sigmoid function.

[0015] Optionally, in the above-mentioned method embodiments of the present invention, the update representation vector of the bidding project of the jth layer is obtained through the embedding propagation layer. .

[0016] Optionally, in the above method embodiments of the present invention, the feature fusion layer assigns different weights to the updated representation vectors of the embedded propagation layers of different layers, and obtains a final representation vector by weighted summing of the updated representation vectors of the embedded propagation layers of different layers, including: The weight of the jth layer is: in, is the weight of the jth layer, N is the number of layers of the embedding propagation layer, is the benchmark parameter; The final representation vector of the user is: in, is the final representation vector of the user, is the weight of the jth layer, is the updated representation vector of the user in the j-th layer of the embedding propagation layer; The final representation vector of the bidding project is: in, is the final representation vector of the user, is the weight of the jth layer, The updated representation vector for the bid items of the j-th layer of the embedding propagation layer.

[0017] Optionally, in the above-mentioned method embodiments of the present invention, the prediction layer uses a two-layer multilayer perceptron network to obtain the prediction score between the user and the bidding item, including: The prediction score is: Among them, W1 and W2 are the weight matrices of the two-layer multilayer perceptron, ReLU() is the ReLU activation function, and are the final representation vectors for users and bidding items, respectively.

[0018] Optionally, in the above-mentioned method embodiments of the present invention, the embedding layer, the embedding propagation layer, the feature fusion layer, and the prediction layer constitute a bidding information recommendation model, and the model parameters are optimized using a loss function, and the loss function is: in, , R + is the observed interaction data set, R - is the unobserved interaction data set, r ui and r uj is the prediction score, σ() is the sigmoid function, λ is the regularization parameter, are trainable model parameters.

[0019] In a second aspect, a bidding information recommendation system is provided, characterized in that it includes: A collection module is used to collect bidding project information and user information, wherein the bidding project information is obtained by crawling the bidding website; The recommendation module includes an embedding layer, an embedding propagation layer, a feature fusion layer, and a prediction layer, where: The embedding layer is used to generate an initial representation vector for each user and bidding item, and input the initial representation vector into the embedding propagation layer; The embedding propagation layer has a multi-layer structure. The embedding propagation layers of different layers learn the structural information of neighbors with different hop counts. The embedding propagation layer assigns weights to neighbor nodes through a multi-head attention mechanism to achieve aggregation of neighbor node information and obtain update representation vectors of user and bidding project nodes. The multi-layer embedding propagation layer outputs multiple update representation vectors, and inputs the multiple update representation features into the feature fusion layer. The feature fusion layer assigns different weights to the updated representation vectors of the embedding propagation layers at different layers, performs weighted summation on the updated representation vectors of the embedding propagation layers at different layers to obtain a final representation vector, and inputs the final representation vector to the prediction layer; The prediction layer uses a two-layer multi-layer perceptron network to obtain the prediction score between the user and the bidding item; The output module is used to select several bidding projects with the highest prediction scores and recommend them to the user.

[0020] In a third aspect, the present invention provides an electronic device, characterized in that the electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the bidding information recommendation method described in the first aspect.

[0021] In a fourth aspect, the present invention provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores computer instructions, and the computer instructions are used to implement the bidding information recommendation method described in the first aspect when executed by a processor.

[0022] Compared with the prior art, the present invention has the following beneficial effects: The bidding information recommendation method, electronic device and readable storage medium provided by the present invention address the problem of bidding information recommendation. A multi-head attention mechanism is used in the embedding propagation layer to enhance the flexibility and accuracy of the embedding representation of users and bidding items. The feature fusion layer assigns different weights to different layers to improve the quality of the overall embedding. A two-layer multilayer perceptron is used to calculate the prediction score to better capture the relationship between the embedding vectors and improve the accuracy of the prediction score.

[0023] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The above and other purposes, features and advantages of the present invention will become more apparent by describing the embodiments of the present invention in more detail in conjunction with the accompanying drawings. The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0025] Figure 1It is a flowchart of a bidding information recommendation method provided by an exemplary embodiment of the present invention.

[0026] Figure 2 It is a structural diagram of a bidding information recommendation system provided by an exemplary embodiment of the present invention. DETAILED DESCRIPTION

[0027] Below, the exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments of the present invention, and it should be understood that the present invention is not limited to the exemplary embodiments described here.

[0028] It should be noted that the relative arrangement of components and steps, the numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention unless specifically stated otherwise.

[0029] Those skilled in the art can understand that the terms "first" and "second" in the embodiments of the present invention are only used to distinguish different steps, devices or modules, etc., and neither represent any specific technical meaning nor indicate the necessary logical order between them.

[0030] It should also be understood that, in the embodiments of the present invention, “plurality” may refer to two or more than two, and “at least one” may refer to one, two or more than two.

[0031] It should also be understood that any component, data or structure mentioned in the embodiments of the present invention can generally be understood as one or more, unless explicitly limited or otherwise indicated in the context.

[0032] Example 1 This embodiment provides a bidding information recommendation method, which can be performed by a bidding information recommendation device. The method is suitable for analyzing bidding project information and user information using a bidding information recommendation system, thereby recommending bidding project information to users. The bidding information recommendation device can be implemented in the form of hardware and / or software, and the bidding information recommendation device can be configured in an electronic device with bidding information recommendation method capabilities.

[0033] like Figure 1 As shown, the method provided in this embodiment includes the following steps: S110, collecting bidding project information and user information, wherein the bidding project information is obtained by crawling the bidding website; Among them, the bidding project is published by the bidding party on the institutional procurement website, industry bidding website or company official website, and the bidding party will provide bidding documents, including bidding announcements, bidding requirements, project specifications, bid evaluation criteria, contract terms, etc. Therefore, crawling bidding information from the website where the bidding party publishes the bidding project can obtain bidding information conveniently and timely.

[0034] S120, generating an initial representation vector for each user and bidding item through an embedding layer; S130, inputting the initial representation vector into a multi-layer embedding propagation layer, wherein the embedding propagation layers of different layers learn structural information of neighbors with different hop counts, and the embedding propagation layer assigns weights to neighbor nodes through a multi-head attention mechanism to achieve aggregation of neighbor node information, thereby obtaining updated representation vectors of user and bidding project nodes, and the multi-layer embedding propagation layer outputs multiple updated representation vectors; S140, inputting the multiple updated representation features into a feature fusion layer, the feature fusion layer assigning different weights to the updated representation vectors of the embedded propagation layers at different layers, and performing weighted summation of the updated representation vectors of the embedded propagation layers at different layers to obtain a final representation vector; S150, inputting the final representation vector into a prediction layer, wherein the prediction layer uses a two-layer multilayer perceptron network to obtain a prediction score between the user and the bidding item; S160: Select several bidding projects with the highest prediction scores and recommend them to the user.

[0035] In step S110, web crawling technology, usually referred to as web crawler or web crawling, refers to a program or script that automatically crawls information (such as web pages, pictures, videos, etc.) on the Internet according to certain rules. This technology is widely used in search engines, data mining, information aggregation, market research and other fields.

[0036] Therefore, in this embodiment, the bidding project information is collected using a bidding information collection system. After the bidding information collection system is configured, the bidding information collection system automatically crawls the latest data from the bidding website every day and extracts the bidding project features. The configuration includes the starting URL of the website, the URL of the subsequent page, the relevant request parameters, the data extraction rules and the time extraction rules.

[0037] The bidding project characteristics include the bidding project number, publishing region, publishing unit, bidding unit, agency unit, publishing time, bidding type, bidding product, amount, and bidding deadline.

[0038] User characteristics are extracted from the user information, and the user characteristics include user basic information, user qualification information, user financial information, user legal litigation information, and user historical bidding information.

[0039] Among them, the user's basic information includes the user's company's business scope, company's registered capital, and company location; the user's qualification information includes the user's qualification level, credit level, and professional qualification personnel information, among which the professional qualification personnel information includes the number, level, and years of experience of professional qualification personnel; the user's financial information includes financial reports in recent years; the user's legal litigation information includes the user's involvement in litigation and arbitration in recent years; the user's historical bidding information includes the user's bidding information and winning project data.

[0040] In step S120, a user ID and a bidding project ID are created for the user and the bidding project respectively, and an embedding matrix is ​​constructed according to the user ID and the bidding project ID.

[0041] Encode user features and bidding project features to build user feature embedding vectors and the bidding item feature embedding vector , R d It is a d-dimensional matrix composed of user and bidding item feature embedding vectors, and l represents the lth feature.

[0042] The user's initial representation vector is , n is the number of users, and the initial representation vector of the bidding project is , m is the number of bidding projects.

[0043] In step S130, using a multi-layer embedding propagation layer can capture higher-order relationships between nodes, rather than just directly connected first-order neighbors, which helps to understand patterns and dependencies in complex network structures. Multi-layer embedding propagation layers can increase the expressive power of the model, enabling it to learn more complex feature combinations. Moreover, multi-layer propagation allows nodes to aggregate information from more distant neighbors, which helps nodes obtain richer feature representations, thereby improving the performance of the model.

[0044] Therefore, in this embodiment, the embedded propagation layer has an N-layer structure, and the j-th embedded propagation layer corresponds to a neighbor j hops away from the entity, where N is an integer greater than 1, j=1, 2, ..., N.

[0045] The embedding propagation layer assigns weights to neighbor nodes through a multi-head attention mechanism to aggregate neighbor node information and obtain updated representation vectors of user and bidding project nodes, including: In the j-th layer of the embedding propagation layer, the set of j-hop neighbor nodes of a user u is , where l is the number of neighboring nodes of i hops, j=1,2,…,N, N is the number of layers of the embedded propagation layer, and N is an integer greater than 1.

[0046] In this embodiment, a multi-head attention mechanism is used to capture information from adjacent neighbor nodes to obtain a richer data representation. By using multiple attention heads in parallel, the model can consider multiple feature combinations at the same time, thereby enhancing the expressiveness and flexibility of the model.

[0047] The weight of user u to neighbor node v1 is calculated through the attention mechanism: Among them, π v1 is the weight of user u to neighbor node v1, u T is the transpose of the vector matrix of user u, e u and e v1 are the initial representation vectors of user u and neighbor node v1 respectively, and W is the transformation matrix; Use the softmax function for normalization, expressed as: in, is the normalized weight, e vk Represents the initial representation vector of the neighbor node vk; The neighbor node set representation of user u is obtained by weighted addition of all neighbor nodes, which is expressed as: Among them, e Nu is the neighbor node set of user u, is the normalized weight, e vk Represents the initial representation vector of the neighbor node vk; Aggregate the initial representation vector of user u and the neighbor node set representation of user u, expressed as: Among them, agg is the aggregation vector, LeakyReLU() is the LeakyReLU activation function, W is the conversion matrix, b is the offset, e u is the initial representation vector of user u, e Nu is the neighbor node set of user u; Under the multi-head attention mechanism, the aggregate vectors obtained under each attention mechanism are concatenated to obtain the updated representation vector of the user of the j-th layer of the embedding propagation layer, which is expressed as: in, is the updated representation vector of the user in the j-th embedding propagation layer, P is the number of heads of the attention mechanism, ⊕ is the concatenation operation, and σ() is the sigmoid function.

[0048] In this embodiment, the tender item node is processed using an embedded propagation layer, and the updated representation vector of the tender item at the jth layer is obtained through the embedded propagation layer. .

[0049] In step S140, since different layers of embedding propagation layers provide different information about users and bidding items, if each layer is set to a uniform weight, although the model is simplified, the influence of different layers on the final prediction structure is ignored. Therefore, in this embodiment, by assigning different weights to different layers of embedding propagation layers, it is ensured that not too much information is lost during the propagation process, and the embedding of different layers is retained, which improves the expressiveness of the embedding and the learning ability of the model, and the overall performance is improved.

[0050] Specifically, the weight of the jth layer is: in, is the weight of the jth layer, N is the number of layers of the embedding propagation layer, is the benchmark parameter; The final representation vector of the user is: in, is the final representation vector of the user, is the weight of the jth layer, is the updated representation vector of the user in the j-th layer of the embedding propagation layer; The final representation vector of the bidding project is: in, is the final representation vector of the user, is the weight of the jth layer, The updated representation vector for the bid items of the j-th layer of the embedding propagation layer.

[0051] In step S150 , a two-layer multilayer perceptron (MLP) model is used, which has a stronger ability to capture the complex relationship between node embedding vectors and is more suitable for rating prediction.

[0052] The prediction score is: Among them, W1 and W2 are the weight matrices of the two-layer multilayer perceptron, ReLU() is the ReLU activation function, and are the final representation vectors for users and bidding items, respectively.

[0053] In this embodiment, the loss function plays a core role in the recommendation system. It not only affects the training effect of the model, but also directly affects the performance of the recommendation system and the user experience. The loss function is used to quantify the difference between the model's predicted value and the actual value. By calculating the loss value, the accuracy of the model's prediction of the interaction between the user and the bidding project can be evaluated.

[0054] The embedding layer, embedding propagation layer, feature fusion layer, and prediction layer constitute a bidding information recommendation model. The model parameters are optimized using a loss function. The loss function is: in, , R + is the observed interaction data set, R - is the unobserved interaction data set, r ui and r uj is the prediction score, σ() is the sigmoid function, λ is the regularization parameter, are trainable model parameters.

[0055] In step S160, for a user, the bidding projects are sorted from high to low according to the predicted scores, and a predetermined number of bidding project information is recommended to the user according to the needs.

[0056] Example 2 This embodiment provides a bidding information recommendation system. Figure 2 As shown, the system includes: The collection module 100 is used to collect bidding project information and user information, wherein the bidding project information is obtained by crawling from the bidding website; The recommendation module 200 includes an embedding layer 210, an embedding propagation layer 220, a feature fusion layer 230 and a prediction layer 240, wherein: The embedding layer 210 is used to generate an initial representation vector for each user and bidding item, and input the initial representation vector to the embedding propagation layer 220; The embedding propagation layer 220 has a multi-layer structure. The embedding propagation layers 220 of different layers learn the structural information of neighbors with different hop counts. The embedding propagation layer 220 assigns weights to neighbor nodes through a multi-head attention mechanism to achieve aggregation of neighbor node information and obtain update representation vectors of user and bidding project nodes. The multi-layer embedding propagation layer 220 outputs multiple update representation vectors, and inputs the multiple update representation features into the feature fusion layer 230. The feature fusion layer 230 assigns different weights to the updated representation vectors of the embedding propagation layer 220 of different layers, performs weighted summation on the updated representation vectors of the embedding propagation layer 220 of different layers to obtain a final representation vector, and inputs the final representation vector to the prediction layer 240; The prediction layer 240 uses a two-layer multilayer perceptron network to obtain the prediction score between the user and the bidding item; The output module 300 is used to select several bidding projects with the highest prediction scores and recommend them to the user.

[0057] Example 3 This embodiment provides an electronic device, characterized in that the electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the bidding information recommendation method described in the first aspect.

[0058] Example 4 This embodiment provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the bidding information recommendation method described in the first aspect when executed.

[0059] The basic principles of the present disclosure are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present disclosure are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present disclosure. In addition, the specific details disclosed above are only for the purpose of illustration and ease of understanding, and are not limitations. The above details do not limit the present disclosure to the necessity of adopting the above specific details to be implemented.

[0060] Each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system embodiment, since it basically corresponds to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0061] The block diagrams of the devices, apparatuses, equipment, and systems involved in this disclosure are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagrams. As will be appreciated by those skilled in the art, these devices, apparatuses, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including," "comprising," "having," and the like are open words, referring to "including but not limited to," and can be used interchangeably therewith. The words "or" and "and" used herein refer to the words "and / or," and can be used interchangeably therewith, unless the context clearly indicates otherwise. The word "such as" used herein refers to the phrase "such as but not limited to," and can be used interchangeably therewith.

[0062] The method and apparatus of the present disclosure may be implemented in many ways. For example, the method and apparatus of the present disclosure may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present disclosure are not limited to the order specifically described above, unless otherwise specifically stated. In addition, in some embodiments, the present disclosure may also be implemented as a program recorded in a recording medium, which includes machine-readable instructions for implementing the method according to the present disclosure. Therefore, the present disclosure also covers a recording medium storing a program for executing the method according to the present disclosure.

[0063] It should also be noted that in the apparatus, equipment and method of the present disclosure, each component or each step can be decomposed and / or recombined. These decompositions and / or recombinations should be regarded as equivalent schemes of the present disclosure. The above description of the disclosed aspects is provided to enable any technician in the field to make or use the present disclosure. Various modifications to these aspects are very obvious to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of the present disclosure. Therefore, the present disclosure is not intended to be limited to the aspects shown here, but to the widest scope consistent with the principles and novel features disclosed herein.

[0064] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present disclosure to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.

Claims

1. A bidding information recommendation method, characterized in that: include: Collecting bidding project information and user information, wherein the bidding project information is obtained by crawling the bidding website; Generate initial representation vectors for each user and bidding item through the embedding layer; The initial representation vector is input into a multi-layer embedding propagation layer. The embedding propagation layers at different layers learn the structural information of neighbors with different hop counts. The embedding propagation layer assigns weights to neighbor nodes through a multi-head attention mechanism to aggregate neighbor node information and obtain updated representation vectors of user and bidding project nodes. The multi-layer embedding propagation layer outputs multiple updated representation vectors. Inputting the multiple updated representation features into a feature fusion layer, the feature fusion layer assigns different weights to the updated representation vectors of the embedding propagation layers at different layers, and performing weighted summation of the updated representation vectors of the embedding propagation layers at different layers to obtain a final representation vector; Inputting the final representation vector into a prediction layer, wherein the prediction layer uses a two-layer multilayer perceptron network to obtain a prediction score between the user and the bidding item; A number of bidding projects with the highest prediction scores are selected and recommended to the user.

2. The method according to claim 1, characterized in that The bidding project information is obtained by crawling from the bidding website, including: The bidding project information is collected using a bidding information collection system. After the bidding information collection system is configured, the bidding information collection system automatically crawls the latest data from the bidding website every day and extracts the characteristics of the bidding project.

3. The method according to claim 2, characterized in that The configuration includes the starting URL of the website, the URL of the subsequent page turning, relevant request parameters, data extraction rules and time extraction rules.

4. The method according to claim 2, characterized in that: The bidding project characteristics include the bidding project number, publishing region, publishing unit, bidding unit, agency unit, publishing time, bidding type, bidding product, amount, and bidding deadline.

5. The method according to claim 1, characterized in that User characteristics are extracted from the user information, and the user characteristics include user basic information, user qualification information, user financial information, user legal litigation information, and user historical bidding information.

6. The method according to claim 5, characterized in that The basic information of the user includes the business scope of the user's company, the registered capital of the company, and the location of the company; The user qualification information includes user qualification level, credit level, and professional qualification personnel information, wherein the professional qualification personnel information includes the number, level, and years of experience of professional qualification personnel; The user's financial information includes financial reports in recent years; The user's legal litigation information includes the user's involvement in litigation and arbitration in recent years; The user's historical bidding information includes the user's bidding information and the winning project data.

7. The method according to claim 1, characterized in that The embedding layer generates an initial representation vector for each user and bidding item, including: Create user IDs and bidding project IDs for users and bidding projects respectively, and build an embedding matrix based on the user IDs and bidding project IDs; Encode user features and bidding project features to build user feature embedding vectors and the bidding item feature embedding vector , R d is a d-dimensional matrix consisting of user and bidding item feature embedding vectors, where l represents the lth feature; The user's initial representation vector is , n is the number of users, and the initial representation vector of the bidding project is , m is the number of bidding projects.

8. A bidding information recommendation system, characterized in that: include: A collection module is used to collect bidding project information and user information, wherein the bidding project information is obtained by crawling the bidding website; The recommendation module includes an embedding layer, an embedding propagation layer, a feature fusion layer, and a prediction layer, where: The embedding layer is used to generate an initial representation vector for each user and bidding item, and input the initial representation vector into the embedding propagation layer; The embedding propagation layer has a multi-layer structure. The embedding propagation layers of different layers learn the structural information of neighbors with different hop counts. The embedding propagation layer assigns weights to neighbor nodes through a multi-head attention mechanism to achieve aggregation of neighbor node information and obtain update representation vectors of user and bidding project nodes. The multi-layer embedding propagation layer outputs multiple update representation vectors, and inputs the multiple update representation features into the feature fusion layer. The feature fusion layer assigns different weights to the updated representation vectors of the embedding propagation layers at different layers, performs weighted summation on the updated representation vectors of the embedding propagation layers at different layers to obtain a final representation vector, and inputs the final representation vector to the prediction layer; The prediction layer uses a two-layer multi-layer perceptron network to obtain the prediction score between the user and the bidding item; The output module is used to select several bidding projects with the highest prediction scores and recommend them to the user.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the bidding information recommendation method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the bidding information recommendation method according to any one of claims 1 to 7 when executed.

Images