A method, apparatus, device, and storage medium for determining feature information

Abstract

The present invention discloses a method, apparatus, device and storage medium for determining feature information. The present invention relates to the field of artificial intelligence. The method includes: reading relationship information between entity elements from a pre-generated entity relationship table; wherein the entity elements include at least one enterprise and at least one service provided by the platform; determining a feature vector of a graph model formed by the relationship information based on each relationship information; obtaining a feature matrix generated based on portrait matrices respectively corresponding to each entity element; wherein each portrait matrix is determined based on attribute information of the corresponding entity element; and determining a feature vector of a target enterprise according to the feature vector of the graph model and the feature matrix. The technical solution of the present invention constructs a panoramic enterprise relationship graph model through the enterprise portraits, enterprise activities and enterprise relationships of customers on the platform, provides interpretable and highly available machine learning features for the recommendation algorithm, and improves the accuracy and recall rate of the recommendation model.

Description

Technical Field

[0001] The embodiments of the present invention relate to the technical field of artificial intelligence, and in particular, to a method, apparatus, device, and storage medium for determining feature information. Background Art

[0002] When using a recommendation algorithm or a recommendation model for a corporate recommendation platform service, generally, the feature information of an enterprise is first obtained based on enterprise portrait data, and then the feature information is input into the recommendation algorithm or the recommendation model, and the result information of the platform service recommended for the enterprise output by the recommendation algorithm or the recommendation model is obtained. The above-mentioned enterprise portrait data includes multi-dimensional enterprise business information data such as the basic situation data and business operation data of the enterprise.

[0003] However, the feature information obtained only based on the enterprise portrait data is relatively single, cannot comprehensively reflect the enterprise features, and the extraction efficiency is also low. Summary of the Invention

[0004] The embodiments of the present invention provide a method, apparatus, medium, and electronic device for determining feature information to improve the extraction efficiency of enterprise feature information and enhance the accuracy and comprehensiveness of the extracted feature information.

[0005] In a first aspect, the embodiments of the present invention provide a method for determining feature information, including:

[0006] Reading the relationship information between entity elements from a pre-generated entity relationship table; wherein, the entity elements include at least one enterprise and at least one service provided by the platform;

[0007] Determining the feature vector of the graph model formed by each of the relationship information based on each of the relationship information;

[0008] Obtaining a feature matrix generated based on the portrait matrices respectively corresponding to each of the entity elements; wherein each of the portrait matrices is determined based on the attribute information of the corresponding entity element;

[0009] Determining the feature vector of the target enterprise according to the feature vector of the graph model and the feature matrix.

[0010] In a second aspect, the embodiments of the present invention further provide a device for determining feature information, and the device includes:

[0011] A relationship information reading module, configured to read the relationship information between entity elements from a pre-generated entity relationship table; wherein, the entity elements include at least one enterprise and at least one service provided by the platform;

[0012] A feature vector determining module, configured to determine the feature vector of the graph model formed by each of the relationship information based on each of the relationship information;

[0013] A feature matrix generation module, configured to obtain a feature matrix generated based on portrait matrices respectively corresponding to each of the entity elements; wherein, each of the portrait matrices is determined based on the attribute information of the corresponding entity element;

[0014] A feature vector determination module for the target enterprise, which determines the feature vector of the target enterprise according to the feature vector of the graph model and the feature matrix.

[0015] In a third aspect, an embodiment of the present invention further provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the method for determining feature information as described in any one of the embodiments of the present invention is implemented.

[0016] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the method for determining feature information as described in any one of the embodiments of the present invention is implemented.

[0017] In the embodiment of the present invention, an entity relationship table is generated, and the relationship information between entity elements is read from the generated entity relationship table; the feature vector of the graph model formed by each relationship information is determined according to the relationship information between each entity element; the portrait matrix corresponding to each entity element is determined according to the attribute information of each entity element, and a feature matrix is generated based on the portrait matrix; the feature vector of the target enterprise is determined by the feature vector of the graph model and the feature matrix. It can be seen that the above solution extracts the feature information of the target enterprise according to the pre-established entity relationship table and the feature matrix pre-generated based on the portrait matrices respectively corresponding to each entity element, which can effectively improve the extraction efficiency of the feature information of the target enterprise and enhance the accuracy and comprehensiveness of the extracted feature information. Description of the Drawings

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required to be used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0019] Figure 1 It is a flowchart of a method for determining feature information provided in Embodiment 1 of the present invention;

[0020] Figure 1a It is a schematic diagram of a graph model of an enterprise in Embodiment 1 of the present invention;

[0021] Figure 2Flow chart of a method for generating an entity relationship table in a method for determining feature information provided in Embodiment 2 of the present invention;

[0022] Figure 3 It is a flow chart of another method for determining feature information provided in Embodiment 3 of the present invention;

[0023] Figure 4 Flow chart of another method for determining feature information provided in Embodiment 4 of the present invention;

[0024] Figure 5 Structure diagram of a feature information determination device provided in Embodiment 5 of the present invention;

[0025] Figure 6 Structure diagram of a computer device provided in Embodiment 6 of the present invention. Detailed implementation manners

[0026] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It can be understood that the specific embodiments described herein are only for explaining the present invention, rather than limiting the present invention. Additionally, it should be noted that for the sake of description, only parts related to the present invention rather than all structures are shown in the accompanying drawings.

[0027] 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, terms such as "first", "second", etc. are only used for distinguishing descriptions, and cannot be understood as indicating or implying relative importance. The acquisition, storage, use, processing, etc. of data in the technical solution of the present application all comply with the relevant regulations of national laws and regulations.

[0028] Embodiment 1

[0029] Figure 1 It is a flow chart of a method for determining feature information provided in Embodiment 1 of the present invention. This embodiment is applicable to the situation of extracting enterprise features based on an enterprise network model. This method can be executed by the feature information determination device in the embodiments of the present invention. The device can be implemented in software and / or hardware manners, such as Figure 1 As shown, the method specifically includes the following steps:

[0030] S110, Read the relationship information between entity elements from a pre-generated entity relationship table; wherein, the entity elements include at least one enterprise and at least one service provided by the platform.

[0031] Among them, the entity elements include at least one enterprise and at least one service provided by the platform. Specifically, the method for obtaining the attribute information of entity elements can be as follows: Obtain the enterprise name, industrial and commercial information, and legal person information entered by the user; Verify the reliability and authenticity of the entered information, and after successful verification, obtain information such as the industry, place of registration, registered capital, and number of employees in the industrial and commercial information; Enterprises can also supplement additional information, such as the number of useful independent intellectual property rights, the number of valid patents, the number of scientific research institutions, the enterprise management system certification situation, the enterprise financial report audit opinion, operating income, main operating income, operating profit, net profit, total assets, current assets, and total liabilities. Each entity element can have its own attribute information.

[0032] Optionally, the attribute information of an enterprise can include at least one of the following: industry, place of registration, registered capital, number of employees, sub - sector, number of useful independent intellectual property rights, number of valid patents, number of scientific research institutions, enterprise management system certification situation, enterprise financial report audit opinion, operating income, operating profit, total assets, current assets, and total liabilities. The attribute information of a service can include at least one of the following: scale, duration, number of participating enterprises, amount involved, and industry involved.

[0033] Among them, the relationships can include the relationships between enterprises and the relationships between enterprises and services. The relationships between enterprises include: co - participation relationship, matchmaking relationship, commercial transaction relationship between the two parties of a contract, procurement and being - procured relationship, cooperation relationship, sales relationship, and entrustment and being - entrusted relationship, etc. For example, if enterprise A is a certain enterprise and enterprise B is a subsidiary or a branch office of enterprise A. Then enterprise A and enterprise B are two entities, and the fact that enterprise B is a subsidiary of enterprise A is the relationship between them. The relationships between enterprises and services include: participating in services provided by the exhibition platform, hosting exhibitions, participating in activities, hosting activities, releasing demands, responding to demands, undertaking demands sent by other enterprises, project matchmaking, using loan products and toolkits, etc. For example, enterprise A held a technology exhibition named B. Then enterprise A and technology exhibition B are two entities, and the fact that A held B is the relationship between them. Among them, the entity - relationship table can be a record table for relationship extraction. Relationship extraction can be understood as extracting the relationships between two or more entities from text records. For example, there is a known event: Enterprise A hired enterprise B to hold exhibition C. Then relationship extraction means extracting the relationships between A, B, and C from the known event: A (hired) B, B (held) C. By extracting and recording the relationships from multiple known events, an entity - relationship table can be generated.

[0034] S120, determine the eigenvector of the graph model formed by each relationship information based on each relationship information.

[0035] After obtaining the relationship information between entity elements, a graph model formed by the relationship information is determined. Among them, the graph model can be a graph composed of points and lines to describe the relationship between entity elements. Among them, the points can represent entity elements, and the lines can represent the relationships between entity elements. As shown in the figure, Figure 1a is a graph model composed of enterprise A, enterprise B, enterprise C, enterprise D, and enterprise E and the relationships between them. The lines in the graph represent the relationships between entity elements. The sparse matrix can be used to describe the geometric relationship of the graph model and geometrize the graph model. The eigenvector of the graph model is obtained by reducing the dimension of the sparse matrix. By way of example, assume Figure 1a can be represented by the sparse matrix w, and w can be:

[0036]

[0037] The values of the matrix start from A until E. Among them, AB, AC, and AD respectively represent the relationships between A and B, C, and D, and their values can be a number greater than 0 and less than 1. This number represents the tightness of the relationship between entity elements, from 0 to 1, where 0 means no relationship and 1 represents the closest relationship. For example, from Figure 1a it can be seen that the relationship between A and itself is the closest. The first row and first column of the matrix represent the relationship between A and itself, so the value of the first row and first column of the matrix is 1. There is no relationship between A and E, and the first row and fifth column of the matrix represent the relationship between A and E, so the value of the first row and fifth column of the matrix is 0. Similarly, the values of other elements in the matrix can be obtained. Further, the eigenvector wr of the graph model can be obtained through the decomposition of the matrix.

[0038] S130, obtain the feature matrix generated based on the portrait matrices corresponding to each entity element; among them, each portrait matrix is determined based on the attribute information of the corresponding entity element.

[0039] Among them, the portrait matrix is obtained using the attribute information of the entity element. By way of example, the portrait matrices of A, B, C, D, and E are calculated as wA, wB, wC, wD, and wE respectively. By transposing the intermediate matrix composed of the portrait matrices of A, B, C, D, and E, the feature matrix [wA wB wC wD wE] can be obtained T .

[0040] S140, determine the eigenvector of the target enterprise according to the eigenvector and feature matrix of the graph model.

[0041] Among them, the target enterprise can be an enterprise specified by the user, and the eigenvector of the target enterprise can be used to represent the attribute characteristics of the target enterprise. By way of example, as described in the examples in S130 and S120: Figure 1aThe eigenvector of the graph model is wr, and the portrait matrices of A, B, C, D, and E are wA, wB, wC, wD, and wE respectively. The eigenmatrix composed of the portrait matrices is [wA wB wC wD wE] T Assume that the target enterprise is A. Then the eigenvector of A can be expressed as: wrA = wr × [wA wB wC wD wE] T .

[0042] Optionally, after determining the eigenvector of the target enterprise based on the eigenvector and eigenmatrix of the graph model, input the eigenvector of the target enterprise into a preset service recommendation model, determine the target service recommended for the target enterprise according to the output result of the preset service recommendation model, and push the relevant information of the target service to the terminal device corresponding to the target enterprise.

[0043] Among them, the recommendation model can analyze user behavior data, establish user preferences, use the feature information of user interests to match items or services, etc., and after screening and filtering through the recommendation algorithm, find the recommended objects that the user may be interested in, and finally recommend them to the user. The recommendation models include content-based recommendation models, collaborative filtering-based recommendation models, and user-based collaborative filtering models, etc. Among them, the content-based recommendation model can find item B with content information similar to A according to the item A that the user is interested in, and thus recommend B to A; the collaborative filtering-based model can find the adjacent set of users or items according to the user's historical behavior, and calculate the user's preference for the item based on this, including recommendation algorithms based on domain, graph, association rules, and knowledge; the user-based collaborative filtering model can recommend items liked by other users with similar interests to the user, and the key of the model is to calculate the interest similarity between two users. Among them, the service recommendation model can be used to recommend a certain service for a certain company. The preset service recommendation model can be a model specified by the user or a service recommendation model adopted according to actual situation requirements. After calculation by the preset service model, the output result of the preset service recommendation model is determined as the target service recommended for the target enterprise. For example, enterprise A has used loan project B many times, enterprise C now needs to use a loan project, and it is known that C is a subsidiary of A, then loan project B is recommended to enterprise C. Among them, C is the target enterprise and B is the target service. Further, the relevant information of the target service is pushed to the terminal device corresponding to the target enterprise. Among them, the relevant information of the target service includes information related to the service such as the service organizer, service cost, service location, and service qualification. For example, when recommending service B to enterprise C, the relevant information of service B is pushed to the terminal device corresponding to the target enterprise. The relevant information of service B includes relevant information such as loan amount, loan source, loan amount, and service location. Terminal devices include devices such as mobile phones and computers.

[0044] By inputting the feature vector of the target enterprise into a preset service recommendation model, determining the target service recommended for the target enterprise according to the output result of the preset service recommendation model, and pushing the relevant information of the target service to the terminal device corresponding to the target enterprise. Information filtering technology can be used to recommend service information that the target enterprise may be interested in, improve the efficiency of service information processing, and enhance the user experience.

[0045] In the technical solution of this embodiment, the relationship information between entity elements is read from a pre-generated entity relationship table; wherein, the entity elements include at least one enterprise and at least one service provided by the platform; the feature vector of the graph model formed by each relationship information is determined based on each relationship information; a feature matrix generated based on the portrait matrices respectively corresponding to each entity element is obtained; wherein, each portrait matrix is determined based on the attribute information of the corresponding entity element; the feature vector of the target enterprise is determined according to the feature vector of the graph model and the feature matrix. It is possible to construct a panoramic enterprise relationship graph model, provide interpretable and highly available machine learning features for the recommendation algorithm, and improve the accuracy and recall rate of the recommendation model.

[0046] Embodiment 2

[0047] Figure 2 It is a flowchart of a method for generating an entity relationship table in a method for determining feature information provided in Embodiment 2 of the present invention. This embodiment refines the method for generating the entity relationship table based on the above embodiment. As Figure 2 shown, the method of this embodiment specifically includes the following steps:

[0048] S210, obtain the service usage record table stored in the data warehouse; wherein, each record in the service usage record table includes a service identifier and an enterprise identifier participating in the service.

[0049] Among them, the data warehouse includes enterprise entity tables, service entity tables, and service usage record tables. Among them, the service usage record tables can be used to generate entity relationship tables. Each record in the service usage record tables includes a service identifier and an enterprise identifier participating in the service. The service identifier includes a service unique code, and the enterprise identifier includes an enterprise code. The attributes of the entity relationship table include associated entities, related entities, relationship types, and relationship tightness values. Among them, the relationship types include the relationship between enterprises and the relationship between enterprises and services. Among them, the relationship between enterprises can be obtained by traversing the service usage record tables and grouping them by the service unique code. Then, the enterprises in the same group are paired up in pairs and recorded as a relationship record. The two enterprises are respectively recorded as the associated entity and the related entity, the relationship type is recorded as the relationship between enterprises, and the relationship tightness value is recorded as a numerical value from 0 to 1. Among them, the relationship tightness value can be calculated according to the degree of tightness of its association. For example, in the service usage record table, service C is found according to a service unique code. There is a record in the usage record table of service C: Enterprise A and Enterprise B jointly held exhibition C, and Enterprise D invested in Enterprise A at the exhibition. Now, the enterprises are paired up in pairs and recorded as relationship records. Then, relationship record 1 is that Enterprise A and Enterprise B have a collaborative relationship. Among them, Enterprise A is the associated entity, Enterprise B is the related entity, the relationship type is the collaborative relationship in the relationship between enterprises, and the relationship tightness value is assumed to be 0.2 (the relationship tightness value can be calculated according to the degree of tightness of its association, and the degree of tightness of the association can be preset by the user or determined according to the actual situation. For example, the relationship between a subsidiary and its parent company is relatively tight, with a value of 0.8, and the relationship of jointly participating in a meeting is set to 0.1). Relationship record 2 is that Enterprise A and Enterprise D have an investment relationship. Among them, Enterprise A is the associated entity, Enterprise D is the related entity, the relationship type is the investment relationship in the relationship between enterprises, and the relationship tightness value is assumed to be 0.3. Among them, the relationship between enterprises and services can be obtained by traversing the service usage record tables. A new relationship data is added for each service usage record. The participating service enterprise is recorded as the associated entity, and the service record is recorded as the related entity. The relationship type is recorded as the relationship between enterprises and services, and the relationship tightness value is recorded as a numerical value from 0 to 1. Among them, the relationship tightness value can be calculated according to the degree of tightness of its association. For example, Enterprise A held a large-scale campus recruitment fair C, and Enterprise B is one of the enterprises participating in recruitment fair C. Then, Enterprise B is the participating service enterprise, that is, the associated entity, and service C is the related entity. The relationship type is recorded as the participation activity relationship in the relationship between enterprises and services, and the relationship tightness value is recorded as 0.3.

[0050] S220. Read each record in the service usage record table. For each read record, generate a first relationship information including an associated entity identifier, an associated entity identifier, an association type, and a relationship tightness value, and store each piece of first relationship information in the entity relationship table. Among them, the associated entity identifier in the first relationship information is the enterprise identifier in the corresponding record, the associated entity identifier in the first relationship information is the service identifier in the corresponding record, the relationship type in the first relationship information is the relationship between the enterprise and the service, and the relationship tightness value in the first relationship information is determined according to the value of the preset attribute of the corresponding service.

[0051] Specifically, each record in the service usage record table includes a service identifier and an enterprise identifier participating in the service. By screening out the records of the relationship information between the enterprise and the service in S210, the first relationship information can be obtained. Store each piece of first relationship information in the entity relationship table. Among them, the first relationship information includes an associated entity, an associated entity, a relationship type, and a relationship tightness value. Among them, the relationship type in the first relationship information is the relationship between the enterprise and the service, and the relationship tightness value in the first relationship information is determined according to the value of the preset attribute of the corresponding service.

[0052] In this solution, optionally, the relationships between the enterprise and the service include at least one of the enterprise participating in the platform exhibition, the enterprise hosting the platform exhibition, the enterprise participating in the platform activity, the enterprise hosting the platform activity, the enterprise publishing demand information through the platform, the enterprise responding to the demand information published by the platform, the enterprise using the project matching service of the platform, the enterprise using the business products of the platform, and the enterprise using the business tools of the platform.

[0053] In this solution, optionally, determine the target range interval where the value of the preset attribute of the corresponding service is located; according to the corresponding relationship between the pre-set range interval and the relationship tightness value, determine the relationship tightness value corresponding to the target range interval, and determine this relationship tightness value as the relationship tightness value in the first relationship information.

[0054] Exemplarily, Enterprise A obtained an investment amount of 100 million yuan by using the matching service C. The target range interval of the value of the preset matching service is 0 - 200 million yuan, then the relationship tightness value can be 1 / 2 (0.5). The relationship tightness value in the first relationship obtained by the above method can more reasonably describe the relationship intimacy between the enterprise and the service, and numericalize its relationship intimacy, which is convenient for subsequent steps to calculate.

[0055] S230: Group each record in the service usage record table according to the service identifier, pair up the enterprises in the same group in pairs, and generate a second relationship information for each pair of enterprises, including the associated entity identifier, the associated entity identifier, the relationship type, and the relationship tightness value, and store each second relationship information in the entity relationship table; where the associated entity identifier in the second relationship information is the identifier of one enterprise in the corresponding enterprise pair, the associated entity identifier in the second relationship information is the identifier of the other enterprise in the corresponding enterprise pair, the relationship type in the second relationship information is the relationship between enterprises, and the relationship tightness value in the second relationship information is determined based on the relevant information of the corresponding enterprise pair.

[0056] Specifically, each record in the service usage record table includes a service identifier and the identifier of the enterprise participating in the service. Grouping each record in the service usage record table according to the service identifier can obtain the second relationship information. Among them, the second relationship information includes the associated entity, the associated entity, the relationship type, and the relationship tightness value. Among them, the relationship type in the second relationship information is the relationship between enterprises, and the relationship tightness value in the second relationship information is determined based on the value of the preset attribute of the corresponding service.

[0057] Optionally, the relationship between enterprises includes at least one of: subsidiary relationship, co-participation relationship, and business transaction contract relationship.

[0058] In this solution, optionally, determine the target enterprise relationship type between the two enterprises in the corresponding enterprise pair; according to the corresponding relationship between the pre-set enterprise relationship type and the relationship tightness value, determine the relationship tightness value corresponding to the target enterprise relationship type, and determine this relationship tightness value as the relationship tightness value in the second relationship information.

[0059] Exemplarily, the business transaction relationship between Enterprise A and Enterprise B is the relationship between the buyer and the seller, and the pre-set intimacy value of the business transaction relationship is 0.6, then the relationship tightness value between A and B is 0.6. Exemplarily, Enterprise A is a subsidiary of Enterprise B, and the pre-set tightness value of the subsidiary type relationship is 0.8, then the relationship tightness value between A and B is 0.8.

[0060] In this solution, optionally, determine the service jointly participated in by the two enterprises in the corresponding enterprise pair, and determine the relationship tightness value in the second relationship information based on the value of the preset attribute of the service jointly participated in by the two enterprises.

[0061] Exemplarily, Enterprise A and Enterprise B jointly participated in an investment project, and among them, A invested 100 million yuan in B. The preset target range interval of the value of the investment project service is 0 billion - 2 billion, then the relationship tightness value between A and B can be 1 / 2 (0.5).

[0062] The relationship tightness values obtained by the above two methods in the second relationship can more reasonably describe the relationship intimacy between enterprises, and numericalize the relationship intimacy value, which is convenient for subsequent steps to perform calculations.

[0063] The technical solution of this embodiment is to obtain the service usage record table stored in the data warehouse; wherein, each record in the service usage record table includes a service identifier and an enterprise identifier participating in the service; read each record in the service usage record table, and for each read record, generate a first relationship information including an associated entity identifier, an associated entity identifier, an association type, and a relationship tightness value, and store each first relationship information in the entity relationship table; wherein, the associated entity identifier in the first relationship information is the enterprise identifier in the corresponding record, the associated entity identifier in the first relationship information is the service identifier in the corresponding record, the relationship type in the first relationship information is the relationship between the enterprise and the service, and the relationship tightness value in the first relationship information is determined according to the value of the preset attribute of the corresponding service; group the records in the service usage record table according to the service identifier, pair the enterprises in the same group in pairs, and generate a second relationship information including an associated entity identifier, an associated entity identifier, an association type, and a relationship tightness value for each pair of enterprises, and store each second relationship information in the entity relationship table; wherein, the associated entity identifier in the second relationship information is the identifier of one enterprise in the corresponding enterprise pair, the associated entity identifier in the second relationship information is the identifier of the other enterprise in the corresponding enterprise pair, the relationship type in the second relationship information is the relationship between enterprises, and the relationship tightness value in the second relationship information is determined according to the relevant information of the corresponding enterprise pair. It can generate an entity relationship information table for the method of determining feature information, improve the calculation efficiency of the method of determining feature information, and reduce the calculation amount of the method of determining feature information.

[0064] Embodiment 3

[0065] Figure 3 It is a flowchart of another method for determining feature information provided by Embodiment 3 of the present invention. This embodiment refines the feature vector of the graph model formed by each relationship information based on the above embodiments. As Figure 3 shown, the method of this embodiment specifically includes the following steps:

[0066] S310, construct an n*n dimensional sparse matrix based on each relationship information; wherein, n is the total number of entity elements, and the value of the element in the i-th row and j-th column of the sparse matrix is the relationship tightness value between the i-th entity element and the j-th entity element, and i and j take values in [1,n].

[0067] Among them, the relationship information includes the relationship between enterprises and the relationship between enterprises and services. Exemplarily, as Figure 1a shown, fromFigure 1a It can be seen that there are five enterprise entities, namely A, B, C, D, and E, among which there are relationships between A, B, C, D, and E. Let n be the total number of entity elements. At this time, n = 5. Based on Figure 1a a 5*5 sparse matrix can be constructed:

[0068]

[0069] The value of the element in the i-th row and j-th column of the sparse matrix is the relationship tightness value between the i-th entity element and the j-th entity element. For example, the value in the second row and third column of the matrix is the relationship value between entity B and entity C.

[0070] S320. Determine the eigenvector of the graph model formed by each relationship information according to the sparse matrix.

[0071] Among them, the eigenvector can reveal the relationship between entities under the condition of dimensionality reduction.

[0072] Optionally, by performing matrix decomposition on the sparse matrix, the sparse matrix is reduced to a one-dimensional vector, and the one-dimensional vector is determined as the eigenvector of the graph model.

[0073] Among them, matrix decomposition can disassemble a matrix into the product of several matrices. Matrix decomposition methods include triangular decomposition, full-rank decomposition, and singular value decomposition, etc. For example, through the matrix decomposition algorithm, the matrix in S310 is reduced to a one-dimensional eigenvector wr. Obtaining the eigenvector of the matrix through matrix decomposition can reduce the calculation amount, save calculation time, and improve the efficiency of the entire model.

[0074] S330. For each entity element, determine the corresponding attribute vector for each attribute information of the current entity element, and generate the portrait matrix corresponding to the current entity element based on the attribute vectors respectively corresponding to each attribute information of the current entity element.

[0075] Among them, each entity can be a service provided by an enterprise or a platform, and the corresponding attribute vector is determined for each attribute information of the current entity element. For example. The attributes of enterprise A include five attributes a, b, c, d, and e, then the eigenvector corresponding to enterprise A is determined according to its attribute information. Further, the portrait matrix corresponding to the current entity element is generated based on the attribute vectors respectively corresponding to each attribute information. For example, based on the eigenvectors corresponding to each attribute, A, B, C, D, and E calculate to obtain the portrait matrices wA, wB, wC, wD, and wE.

[0076] S340. Generate an intermediate matrix containing the portrait matrices respectively corresponding to each entity element, and determine the transposed matrix of the intermediate matrix as the eigenmatrix.

[0077] Specifically, an intermediate matrix is generated based on the portrait matrices respectively corresponding to each entity element. For example, an intermediate matrix [wA wB wC wD wE] can be generated through each portrait matrix wA, wB, wC, wD, and wE. Further, [wA wBwC wD wE] T is determined as the feature matrix.

[0078] S350, calculate the product result of the feature vector of the graph model and the feature matrix, and determine the product result as the feature vector of the target enterprise.

[0079] Among them, the target enterprise can be the enterprise specified by the user, and the feature vector of the target enterprise can be used to represent the attribute information of the enterprise, the relationship between the enterprise and other enterprises, and the relationship between the enterprise and each service. By way of example, as described in the above example: Figure 1a The feature vector is wr, and the feature matrix is [wA wB wC wD wE] T . Assuming the target enterprise is A, the feature vector of A can be expressed as: wrA = wr × [wA wB wC wD wE] T .

[0080] The technical solution of this embodiment constructs an n*n dimensional sparse matrix based on each relationship information; where n is the total number of entity elements, and the value of the element in the i-th row and j-th column of the sparse matrix is the relationship tightness value between the i-th entity element and the j-th entity element, and i and j take values in [1, n]; determine the feature vector of the graph model formed by each relationship information according to the sparse matrix; for each entity element, determine the corresponding attribute vector for each piece of attribute information of the current entity element, and generate the portrait matrix corresponding to the current entity element based on the attribute vectors respectively corresponding to each piece of attribute information of the current entity element; generate an intermediate matrix including the portrait matrices respectively corresponding to each entity element, and determine the transpose matrix of the intermediate matrix as the feature matrix; calculate the product result of the feature vector of the graph model and the feature matrix, and determine the product result as the feature vector of the target enterprise. It can construct a panoramic enterprise relationship graph model, provide interpretable and highly available machine learning features for the recommendation algorithm, and improve the accuracy and recall rate of the recommendation model.

[0081] Embodiment 4

[0082] Figure 4 This is a flowchart of another method for determining feature information provided by Embodiment 4 of the present invention. This embodiment refines the steps after determining the feature vector of the target enterprise based on the feature vector and feature matrix of the graph model on the basis of the above embodiments. As Figure 4 shown, the method of this embodiment specifically includes the following steps:

[0083] S410, obtain the determined weight vector; wherein, the weight vector includes the adjusted weight values corresponding to the respective attribute information of each entity element.

[0084] Among them, weight can be understood as the importance degree of a certain factor or index relative to a certain thing. It is different from the general proportion. It reflects not only the percentage of a certain factor or index, but emphasizes the relative importance degree of the factor or index, tending to the contribution degree or importance. The weight value can be the value of the weight. The weight values of each element represent the size of each element in the overall and the importance degree of influencing the overall ability. By way of example, the number of times enterprise A used service B in the past year was 10 times, and the number of times enterprise A used service C in the past year was 5 times. Then, in the relevant information of enterprise A, the weight value of service B is larger than the weight value of service C. Because enterprise A used service B more times, so service B is more "important" to enterprise A than service C. Among them, the weight vector can be a vector composed of the adjusted weight values corresponding to the respective attribute information of each entity element.

[0085] In this solution, optionally, obtaining the determined weight vector can be achieved by obtaining the recommended feedback information corresponding to each service; the recommended feedback information includes whether each enterprise uses the corresponding recommended service; and determining the weight vector according to the recommended feedback information.

[0086] Determine the weight vector according to the recommended feedback information. By way of example, it includes the following steps:

[0087] Step 1, define a one-dimensional vector wt with m elements as the initial weight vector: wt = [1, 1,... 1]. Among them, the value of m is the number of entity elements. Define the error threshold sigma and the convergence step size step. Among them, the error threshold can be a scalar value standard for separating positive and negative categories; among them, convergence can be that each entity passes the calculation of the model, and the discrete errors of each node representing the entity on a certain function used by the model all approach 0; among them, the convergence step size can be used to control the size of the change amount of each weight value in each step; assume that the currently traversed i-th service is service C. For service C, calculate the existing recommended feedback of the enterprise, and the existing feedback recommendations can be divided into two groups. If the enterprise accesses or uses this service, record the recommended feedback as 1; if the enterprise does not access or use this service, record the recommended feedback as -1.

[0088] Step 2: Randomly select two enterprise recommended feedbacks from the two groups respectively or randomly select two groups of enterprises to calculate the mean value of the recommended feedbacks of the two groups of enterprises. Assume that the recommended feedback values of the two selected enterprises are enterprise A and enterprise B. Denote the feature vectors of enterprise A and enterprise B as wA and wB.

[0089] Step 3: Calculate the gradient lumbda(0) of vectors wA and wB. Among them, the gradient can represent the directional derivative of a certain function at a certain point, which reaches the maximum along this direction, that is, the function changes fastest and has the largest change rate at this point along this direction (the direction of this gradient). The methods for calculating the gradient include numerical methods, analytical methods, backpropagation methods, etc.

[0090] Step 4: Update the current wt. The updated wt = step * lumbda(0) * the previous wt. Calculate the sum of the distances from the existing recommendations to the updated wt. If sum is greater than or equal to sigma, then generalize by mathematical induction and repeat Steps 2 to 4; if sum is less than sigma, then take the current wt as the final weight vector of the current service i, that is, wt(i).

[0091] Step 5: Similar to Steps 1 to 4, traverse other services and repeat Steps 1 to 4 to obtain the final weight vector wt(i) for each service, where i takes values in [1, N], and N is the total number of services.

[0092] Step 6: Assume that v(i) is the access volume of service i. Among them, the access volume can be the number of times the service is used, then the total access volume of the services v = v(1) + v(2)... + v(N). The final required weight vector can be obtained by allocating weight ratios according to the service access volume. Denote the obtained final weight vector as wt(O), then the calculation formula of wt(O) can be: wt(O) = v(1) / v * wt(1) + v(2) / v * wt(2) +... + v(N) / v * wt(N).

[0093] By determining the weight vector according to the recommendation feedback information, relevant services can be recommended to users according to the user's behavior habits and preferences, improving the user experience.

[0094] S420. Adjust the feature vector of the target enterprise based on the weight vector.

[0095] As in the above steps, the final weight vector obtained is wt. Adjust the feature vector of the target enterprise based on the weight vector. Assume that the target enterprise is Enterprise A, and the adjusted feature vector of the target enterprise is W, then W = wt * wrA. After obtaining the feature vector, this feature vector can be used as the input of any recommendation model, such as support vector machines (SVM), clustering analysis, deep learning, etc.

[0096] The technical solution of this embodiment is to obtain a determined weight vector, where the weight vector contains the adjustment weight values corresponding to the respective attribute information of each entity element, and adjust the feature vector of the target enterprise based on the weight vector. It can recommend relevant services to users according to the user's behavior habits and preferences, improve the accuracy and recall rate of the recommendation model, and further enhance the user experience.

[0097] Embodiment 5

[0098] Figure 5 FIG. 5 is a schematic structural diagram of a device for determining feature information provided by Embodiment 5 of the present invention. This embodiment is applicable to the case of calculating features based on an enterprise network model. The device can be implemented in software and / or hardware, and can be integrated in any device that provides the function of determining feature information, such as Figure 5 As shown, the device for determining feature information specifically includes:

[0099] A relationship information reading module 510, configured to read the relationship information between entity elements from a pre-generated entity relationship table, where the entity elements include at least one enterprise and at least one service provided by the platform;

[0100] A feature vector determining module 520, configured to determine the feature vector of the graph model formed by each relationship information based on each relationship information;

[0101] A feature matrix generating module 530, configured to obtain a feature matrix generated based on the portrait matrices corresponding to each entity element, where each portrait matrix is determined based on the attribute information of the corresponding entity element;

[0102] A feature vector determining module 540 of the target enterprise, configured to determine the feature vector of the target enterprise according to the feature vector of the graph model and the feature matrix.

[0103] The device for determining feature information provided by Embodiment 5 of the present invention reads the relationship information between entity elements from a pre-generated entity relationship table, where the entity elements include at least one enterprise and at least one service provided by the platform; determines the feature vector of the graph model formed by each relationship information based on each relationship information; obtains a feature matrix generated based on the portrait matrices corresponding to each entity element, where each portrait matrix is determined based on the attribute information of the corresponding entity element; and determines the feature vector of the target enterprise according to the feature vector of the graph model and the feature matrix. It can construct a panoramic enterprise relationship graph model, provide interpretable and highly available machine learning features for the recommendation algorithm, and improve the accuracy and recall rate of the recommendation model.

[0104] Optionally, the feature vector determining module 540 of the target enterprise is further configured to:

[0105] Input the feature vector of the target enterprise into the preset service recommendation model, determine the target service recommended for the target enterprise according to the output result of the preset service recommendation model, and push the relevant information of the target service to the terminal device corresponding to the target enterprise.

[0106] Optionally, the device further includes:

[0107] Service usage record table acquisition unit: Acquire the service usage record table stored in the data warehouse; where each record in the service usage record table includes a service identifier and an enterprise identifier participating in the service;

[0108] First relationship information generation unit: Read each record in the service usage record table, and generate a piece of first relationship information including an associated entity identifier, an associated entity identifier, an association type, and a relationship tightness value for each read record, and store each piece of first relationship information in the entity relationship table; where the associated entity identifier in the first relationship information is the enterprise identifier in the corresponding record, the associated entity identifier in the first relationship information is the service identifier in the corresponding record, the association type in the first relationship information is the relationship between the enterprise and the service, and the relationship tightness value in the first relationship information is determined according to the value of the preset attribute of the corresponding service;

[0109] Second relationship information generation unit: Group the records in the service usage record table according to the service identifier, pair up the enterprises in the same group, and generate a piece of second relationship information including an associated entity identifier, an associated entity identifier, an association type, and a relationship tightness value for each pair of enterprises, and store each piece of second relationship information in the entity relationship table; where the associated entity identifier in the second relationship information is the identifier of one enterprise in the corresponding enterprise pair, the associated entity identifier in the second relationship information is the identifier of the other enterprise in the corresponding enterprise pair, the association type in the second relationship information is the relationship between enterprises, and the relationship tightness value in the second relationship information is determined according to the relevant information of the corresponding enterprise pair.

[0110] Optionally, the first relationship information generation unit is specifically used for:

[0111] Determine the target range interval where the value of the preset attribute of the corresponding service is located;

[0112] According to the pre-set correspondence between the range interval and the relationship tightness value, determine the relationship tightness value corresponding to the target range interval, and determine the relationship tightness value as the relationship tightness value in the first relationship information.

[0113] Optionally, the second relationship information generation unit is specifically used for:

[0114] Determine the target enterprise relationship type between the two enterprises in the corresponding enterprise pair; according to the correspondence between the preset enterprise relationship type and the relationship tightness value, determine the relationship tightness value corresponding to the target enterprise relationship type, and determine the relationship tightness value in the second relationship information as this relationship tightness value; or,

[0115] Determine the services jointly participated by the two enterprises in the corresponding enterprise pair, and determine the relationship tightness value in the second relationship information according to the value of the preset attribute of the services jointly participated by the two enterprises.

[0116] Optionally, the feature vector determination module 520 includes:

[0117] Sparse matrix construction sub-unit: construct an n*n dimensional sparse matrix based on each relationship information; where n is the total number of entity elements, and the value of the element in the i-th row and j-th column of the sparse matrix is the relationship tightness value between the i-th entity element and the j-th entity element, and i and j take values in [1,n];

[0118] Feature vector formation sub-unit: determine the feature vector of the graph model formed by each relationship information according to the sparse matrix.

[0119] Optionally, the feature vector formation sub-unit is specifically used for:

[0120] By performing matrix decomposition on the sparse matrix, reduce the sparse matrix to a one-dimensional vector, and determine the one-dimensional vector as the feature vector of the graph model.

[0121] Optionally, the feature matrix generation module 530 is further used for:

[0122] For each entity element, determine the corresponding attribute vector for each piece of attribute information of the current entity element, and generate the portrait matrix corresponding to the current entity element based on the attribute vectors respectively corresponding to each piece of attribute information of the current entity element;

[0123] Generate an intermediate matrix containing the portrait matrices respectively corresponding to each entity element, and determine the transposed matrix of the intermediate matrix as the feature matrix.

[0124] Optionally, the feature vector determination module 540 of the target enterprise is specifically used for:

[0125] Calculate the product result of the feature vector of the graph model and the feature matrix, and determine the product result as the feature vector of the target enterprise.

[0126] Optionally, the device further includes:

[0127] Weight vector acquisition unit: acquire the determined weight vector; where the weight vector contains the adjusted weight values respectively corresponding to each piece of attribute information of each entity element;

[0128] Feature vector adjustment unit: Adjust the feature vector of the target enterprise based on the weight vector.

[0129] Optionally, the weight vector acquisition unit is specifically configured to:

[0130] Obtain the recommended feedback information corresponding to each service; the recommended feedback information includes whether each enterprise uses the corresponding recommended service;

[0131] Determine the weight vector according to the recommended feedback information. Optionally, the relationship between enterprises includes at least one of: subsidiary relationship, co - participation relationship, and business transaction contract relationship; Optionally, the relationship between an enterprise and a service includes at least one of: the enterprise participates in a platform exhibition, the enterprise hosts a platform exhibition, the enterprise participates in a platform activity, the enterprise hosts a platform activity, the enterprise publishes demand information through the platform, the enterprise responds to the demand information published by the platform, the enterprise uses the project matchmaking service of the platform, the enterprise uses the business products of the platform, and the enterprise uses the business tools of the platform.

[0132] Optionally, the attribute information of an enterprise includes at least one of: industry, place of registration, registered capital, number of employees, subdivision field, number of useful independent intellectual property rights, number of valid patents, number of scientific research institutions, enterprise management system certification situation, enterprise financial report audit opinion, operating income, operating profit, total assets, current assets, and total liabilities;

[0133] The attribute information of a service includes at least one of: scale, duration, number of participating enterprises, amount involved, and industry involved.

[0134] The above - mentioned product can execute the method provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the method.

[0135] Embodiment Six

[0136] Figure 6 It is a schematic structural diagram of a computer device provided by Embodiment Six of the present invention. Figure 6 It shows a block diagram of an exemplary computer device 12 suitable for implementing the embodiments of the present invention. Figure 6 The shown computer device 12 is only an example, and should not bring any limitation to the functions and usage scope of the embodiments of the present invention.

[0137] As Figure 6 shown, the computer device 12 is presented in the form of a general - purpose computing device. The components of the computer device 12 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 connecting different system components (including the system memory 28 and the processing unit 16).

[0138] Bus 18 represents one or more of several types of bus architectures, including a memory bus or memory controller, a peripheral bus, an Accelerated Graphics Port, a processor bus, or a local bus using any of a variety of bus architectures. By way of example, such architectures include, but are not limited to, Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MAC) bus, Enhanced ISA bus, Video Electronics Standards Association (VESA) local bus, and Peripheral Component Interconnect (PCI) bus.

[0139] Computer device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by computer device 12, including both volatile and nonvolatile media, removable and non-removable media.

[0140] System memory 28 can include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. Computer device 12 can further include other removable / non-removable, volatile / nonvolatile computer system storage media. By way of example only, storage system 34 can be used for reading from and writing to non-removable, nonvolatile magnetic media ( Figure 6 not shown and typically called a "hard disk drive"). Although Figure 6 not shown in the figures, a disk drive for reading from and writing to a removable, nonvolatile magnetic disk (e.g., a "floppy disk"), and an optical disk drive for reading from and writing to a removable, nonvolatile optical disk (e.g., a CD-ROM, a DVD-ROM, or other optical media) can be provided. In these cases, each drive can be connected to bus 18 by one or more data media interfaces. Memory 28 can include at least one program product having a set (e.g., at least one) of program modules that are configured to carry out the functions of embodiments of the present invention.

[0141] A program / utility 40 having a set (at least one) of program modules 42 can be stored, for example, in memory 28, and such program modules 42 include - but are not limited to - an operating system, one or more application programs, other program modules, and program data, each of which examples or some combination thereof may include an implementation of a networking environment. Program modules 42 generally carry out the functions and / or methods of the embodiments described herein.

[0142] The computer device 12 can also communicate with one or more external devices 14 (such as keyboards, pointing devices, display 24, etc.), and can also communicate with one or more devices that enable users to interact with the computer device 12, and / or communicate with any device that enables the computer device 12 to communicate with one or more other computing devices (such as network cards, modems, etc.). This communication can be carried out through the input / output (I / O) interface 22. In addition, in the computer device 12 of this embodiment, the display 24 does not exist as an independent entity, but is embedded in the mirror. When the display surface of the display 24 is not displaying, the display surface of the display 24 and the mirror visually merge into one. Also, the computer device 12 can also communicate with one or more networks (such as local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) through the network adapter 20. As shown in the figure, the network adapter 20 communicates with other modules of the computer device 12 through the bus 18. It should be understood that although not shown in the figure, other hardware and / or software modules can be used in combination with the computer device 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0143] The processing unit 16 executes various functional applications and data processing by running the programs stored in the system memory 28. For example, it implements a method for determining feature information provided by an embodiment of the present invention: reading the relationship information between entity elements from a pre-generated entity relationship table; where the entity elements include at least one enterprise and at least one service provided by the platform; determining the feature vector of the graph model formed by each relationship information based on each relationship information; obtaining a feature matrix generated based on the portrait matrices respectively corresponding to each entity element; where each portrait matrix is determined based on the attribute information of the corresponding entity element; determining the feature vector of the target enterprise according to the feature vector of the graph model and the feature matrix.

[0144] Embodiment Seven

[0145] Embodiment Seven of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, it implements a method for determining feature information provided by all embodiments of the present invention: reading the relationship information between entity elements from a pre-generated entity relationship table; where the entity elements include at least one enterprise and at least one service provided by the platform; determining the feature vector of the graph model formed by each relationship information based on each relationship information; obtaining a feature matrix generated based on the portrait matrices respectively corresponding to each entity element; where each portrait matrix is determined based on the attribute information of the corresponding entity element; determining the feature vector of the target enterprise according to the feature vector of the graph model and the feature matrix.

[0146] Any combination of one or more computer-readable media may be employed. The computer-readable media may be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the computer-readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In this document, a computer-readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

[0147] A computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which the computer-readable program code is carried. Such a propagated data signal may take any of a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. The computer-readable signal medium may also be any computer-readable medium that is not a computer-readable storage medium and that can send, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.

[0148] The program code embodied on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0149] The computer program code for carrying out operations of the present invention may be written in one or more programming languages, or combinations thereof, including object-oriented programming languages such as Java, Smalltalk, C++, and also including conventional procedural programming languages such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer, or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (for example, through the Internet using an Internet service provider).

[0150] Note that the above is only the preferred embodiment of the present invention and the technical principles applied. Those skilled in the art will understand that the present invention is not limited to the specific embodiments here, and various obvious changes, re-adjustments and substitutions can be made by those skilled in the art without departing from the protection scope of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments only. Without departing from the concept of the present invention, more other equivalent embodiments can be included, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A method for determining characteristic information, characterized in that The method includes: Reading the relationship information between entity elements from a pre-generated entity relationship table; wherein, the entity elements include at least one enterprise and at least one service provided by the platform; Determining the feature vector of the graph model formed by each of the relationship information based on each of the relationship information; wherein, the relationship information includes the relationship between enterprises and the relationship between enterprises and services; Obtaining a feature matrix generated based on the portrait matrices respectively corresponding to each of the entity elements; wherein, each of the portrait matrices is determined based on the attribute information of the corresponding entity element; the attribute information of the enterprise includes at least one of: industry, place of registration, registered capital, number of employees, sub-sector, number of useful independent intellectual property rights, number of valid patents, number of scientific research institutions, enterprise management system certification situation, enterprise financial report audit opinion, operating income, operating profit, total assets, current assets, and total liabilities; the attribute information of the service includes at least one of: scale, duration, number of participating enterprises, amount involved, and industry involved; Determining the feature vector of the target enterprise according to the feature vector of the graph model and the feature matrix; Obtaining the determined weight vector; wherein, the weight vector contains the adjusted weight values respectively corresponding to the attribute information of each entity element; adjusting the feature vector of the target enterprise based on the weight vector; the method for determining the weight vector includes: obtaining the recommended feedback information corresponding to each service; the recommended feedback information includes whether each enterprise uses the recommended corresponding service; determining the weight vector according to the recommended feedback information; The determining the weight vector according to the recommended feedback information includes: for each service, defining a one-dimensional vector wt with m elements as the initial weight vector: wt = [1, 1,... 1]; wherein, the value of m is the number of entity elements, defining an error threshold sigma and a convergence step size step; wherein the error threshold is a scalar value standard for separating positive and negative categories; randomly selecting two enterprise recommended feedbacks or randomly selecting two groups of enterprises from the recommended feedback information to calculate the mean of the recommended feedbacks of the two groups of enterprises; calculating the gradient lumbda(0) of the feature vectors of the two enterprises; updating wt according to the formula: updated wt = step * lumbda(0) * previous wt; calculating the sum of the distances from the pre-determined existing recommended feedback to the updated wt; if the sum is greater than or equal to sigma, then repeat the step of randomly selecting two enterprise recommended feedbacks or randomly selecting two groups of enterprises from the recommended feedback information to calculate the mean of the recommended feedbacks of the two groups of enterprises; if the sum is less than sigma, then determine the updated wt as the weight vector wt(i) of the current service i; wherein, wt represents the weight vector, wt(i) represents the weight vector of service i; wt and wt(i) are the same parameter; Defining N as the total number of services and v(i) as the access volume of service i; wherein, the access volume is the number of times service i is used, and determining the weight vector according to the following formula: wt(O) = v(1) / v * wt(1) + v(2) / v * wt(2) +... + v(N) / v * wt(N); where wt(O) is the weight vector, and v is the total number of accesses to the service.

2. The method according to claim 1, characterized in that After determining the eigenvector of the target enterprise based on the eigenvector and the eigenmatrix of the graph model, the method further includes: Inputting the eigenvector of the target enterprise into a preset service recommendation model, determining the target service recommended for the target enterprise according to the output result of the preset service recommendation model, and pushing the relevant information of the target service to the terminal device corresponding to the target enterprise.

3. The method according to claim 1, wherein The method for generating the entity relationship table includes: Obtaining a service usage record table stored in a data warehouse; each record in the service usage record table includes a service identifier and an enterprise identifier participating in the service. Reading each record in the service usage record table, generating, for each read record, a first relationship information including an associated entity identifier, an associated entity identifier, an association type, and a relationship tightness value, and storing each piece of the first relationship information in the entity relationship table; where the associated entity identifier in the first relationship information is the enterprise identifier in the corresponding record, the associated entity identifier in the first relationship information is the service identifier in the corresponding record, the relationship type in the first relationship information is the relationship between an enterprise and a service, and the relationship tightness value in the first relationship information is determined based on the value of a preset attribute of the corresponding service. Grouping the records in the service usage record table according to the service identifier, pairing the enterprises in the same group in pairs, and generating, for each pair of enterprises, a second relationship information including an associated entity identifier, an associated entity identifier, an association type, and a relationship tightness value, and storing each piece of the second relationship information in the entity relationship table; where the associated entity identifier in the second relationship information is the identifier of one enterprise in the corresponding pair of enterprises, the associated entity identifier in the second relationship information is the identifier of the other enterprise in the corresponding pair of enterprises, the relationship type in the second relationship information is the relationship between enterprises, and the relationship tightness value in the second relationship information is determined based on the relevant information of the corresponding pair of enterprises.

4. The method according to claim 3, wherein Determining the relationship tightness value in the first relationship information based on the value of a preset attribute of the corresponding service includes: Determining the target range interval where the value of the preset attribute of the corresponding service is located; According to the pre-set correspondence between the range interval and the relationship tightness value, determining the relationship tightness value corresponding to the target range interval, and determining this relationship tightness value as the relationship tightness value in the first relationship information.

5. The method according to claim 3, characterized in that, Determining the relationship tightness value in the second relationship information based on the relevant information of the corresponding pair of enterprises includes: Determining the target enterprise relationship type between the two enterprises in the corresponding pair of enterprises; according to the pre-set correspondence between the enterprise relationship type and the relationship tightness value, determining the relationship tightness value corresponding to the target enterprise relationship type, and determining this relationship tightness value as the relationship tightness value in the second relationship information; or Determine the services jointly participated by the two enterprises in the corresponding enterprise pair, and determine the relationship tightness value in the second relationship information according to the value of the preset attribute of the services jointly participated by the two enterprises.

6. The method according to claim 3, wherein The determining of the feature vector of the graph model formed by each of the relationship information includes: Construct an n*n dimensional sparse matrix based on each of the relationship information; where n is the total number of entity elements, and the value of the element in the i-th row and j-th column of the sparse matrix is the relationship tightness value between the i-th entity element and the j-th entity element, and i and j take values in [1, n]; Determine the feature vector of the graph model formed by each of the relationship information according to the sparse matrix.

7. The method according to claim 6, wherein The determining of the feature vector of the graph model formed by each of the relationship information according to the sparse matrix includes: By performing matrix decomposition on the sparse matrix, reduce the sparse matrix to a one-dimensional vector, and determine the one-dimensional vector as the feature vector of the graph model.

8. The method according to claim 1, wherein The method for generating the feature matrix includes: For each of the entity elements, determine a corresponding attribute vector for each piece of attribute information of the current entity element, and generate a portrait matrix corresponding to the current entity element based on the attribute vectors respectively corresponding to each piece of attribute information of the current entity element; Generate an intermediate matrix including the portrait matrices respectively corresponding to each of the entity elements, and determine the transpose matrix of the intermediate matrix as the feature matrix.

9. The method according to claim 1, characterized in that, The determining of the feature vector of the target enterprise according to the feature vector of the graph model and the feature matrix includes: Calculate the product result of the feature vector of the graph model and the feature matrix, and determine the product result as the feature vector of the target enterprise.

10. The method according to claim 3, characterized in that, The relationship between enterprises includes at least one of: subsidiary relationship, co-participation relationship, and business transaction contract relationship; The relationship between an enterprise and a service includes at least one of: an enterprise participates in a platform exhibition, an enterprise hosts a platform exhibition, an enterprise participates in a platform activity, an enterprise hosts a platform activity, an enterprise publishes demand information through the platform, an enterprise responds to the demand information published by the platform, an enterprise uses the project matchmaking service of the platform, an enterprise uses the business products of the platform, and an enterprise uses the business tools of the platform.

11. An apparatus for determining feature information, characterized in that Including: A relationship information reading module, configured to read the relationship information between entity elements from a pre-generated entity relationship table; where the entity elements include at least one enterprise and at least one service provided by the platform; A feature vector determining module, configured to determine the feature vector of the graph model formed by each of the relationship information based on each of the relationship information; where the relationship information includes the relationship between enterprises and the relationship between an enterprise and a service; A feature matrix generation module, configured to obtain a feature matrix generated based on portrait matrices respectively corresponding to each of the entity elements; wherein, each of the portrait matrices is determined based on the attribute information of the corresponding entity element; the attribute information of the enterprise includes at least one of: industry, place of registration, registered capital, number of employees, sub - field, number of useful independent intellectual property rights, number of valid patents, number of scientific research institutions, enterprise management system certification situation, enterprise financial report audit opinion, operating income, operating profit, total assets, current assets, and total liabilities; the attribute information of the service includes at least one of: scale, duration, number of participating enterprises, amount involved, and involved industry. A target enterprise feature vector determination module, configured to determine a feature vector of the target enterprise according to the feature vector of the graph model and the feature matrix. Obtain the determined weight vector; wherein, the weight vector contains adjustment weight values respectively corresponding to each attribute information of each entity element; adjust the feature vector of the target enterprise based on the weight vector; the method for determining the weight vector includes: obtaining the recommended feedback information corresponding to each service; the recommended feedback information includes whether each enterprise uses the recommended corresponding service; determine the weight vector according to the recommended feedback information. The determination of the weight vector according to the recommended feedback information includes: for each service, define a one - dimensional vector wt with m elements as the initial weight vector: wt = [1, 1,... 1]; where the value of m is the number of entity elements, define the error threshold sigma and the convergence step size step; where the error threshold is a scalar value standard for separating positive and negative categories; randomly select two enterprise recommended feedbacks or randomly select two groups of enterprises from the recommended feedback information to calculate the mean of the two groups of enterprise recommended feedbacks; calculate the gradient lumbda(0) of the feature vectors of the two enterprises; update wt according to the formula: updated wt = step * lumbda(0) * previous wt; calculate the sum of the distances from the pre - determined existing recommended feedback to the updated wt; if sum is greater than or equal to sigma, repeat the step of randomly selecting two enterprise recommended feedbacks or randomly selecting two groups of enterprises from the recommended feedback information to calculate the mean of the two groups of enterprise recommended feedbacks; if sum is less than sigma, determine the updated wt as the weight vector wt(i) of the current service i; where wt represents the weight vector and wt(i) represents the weight vector of service i; wt and wt(i) are the same parameter. Define N as the total number of services and v(i) as the access volume of service i; where the access volume is the number of times service i is used, and determine the weight vector according to the following formula: wt(O) = v(1) / v * wt(1)+v(2) / v * wt(2)+...+v(N) / v * wt(N); where wt(O) is the weight vector and v is the total access volume of the services.

12. A computer device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the computer program, the method for determining the characteristic information according to any one of claims 1-10 is implemented.

13. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, the method for determining the characteristic information according to any one of claims 1-10 is implemented.

Images