User intelligent matching method and device, equipment and medium

Abstract

The invention discloses a user intelligent matching method and device, equipment and a medium. The method comprises the following steps: generating a first user service feature vector and a second user attribute matrix in a service system through a service system server, and establishing a local graph model; sampling in the local graph model according to the connection relationship between the first user and the second user through the service system server to obtain a plurality of sub-graphs, and training according to the sub-graphs and the global initial graph model to obtain local model parameters; global model parameters are calculated through the central server according to the local model parameters uploaded by the service system servers and the number of registered users of the service systems; and obtaining a user matching result through the service system server according to the currently stored global model parameters. By adopting the technical scheme, the model parameters can be optimized in combination with the data of the multiple service systems on the premise of ensuring the data security of each service system, and the user matching accuracy is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and in particular to a user intelligent matching method and device, equipment and medium. BACKGROUND

[0002] In some business scenarios, it is necessary to improve service efficiency and business conversion effect through accurate personnel matching, and the traditional matching mode relying on human experience has been difficult to meet the needs of efficient operation of the industry.

[0003] At present, the matching is generally performed by a traditional fixed rule static allocation method or a dynamic matching method based on machine learning. The traditional fixed rule static allocation method relies on preset rules to complete the division, and the dynamic matching method based on machine learning realizes dynamic optimization of matching weights by constructing a prediction model, and then obtains a matching result according to the characteristics and the weighted weights.

[0004] However, the traditional fixed rule allocation method has poor dynamic adaptability and cannot optimize the matching rules in real time. Each rule iteration needs to be manually updated by technical personnel, which not only consumes time and effort, but also easily leads to a lag of the matching strategy behind the business needs. In addition, the matching accuracy of this method is low, which easily leads to a decrease in business execution efficiency. As for the allocation method based on machine learning, it is easily limited by industry compliance, and the data of each business system is difficult to realize direct interconnection. The actual data fusion ratio is insufficient, which easily leads to a limited coverage of model training data and significantly affects the prediction accuracy. SUMMARY

[0005] The present application provides a user intelligent matching method, device, equipment and medium, which can optimize model parameters in combination with multi-business system data under the premise of ensuring the safety of data of each business system, and improve the accuracy of user matching.

[0006] According to an aspect of the present application, a user intelligent matching method is provided, which is executed by a central server and a plurality of business system servers, comprising:

[0007] The business system server generates a first user business feature vector and a second user attribute matrix in the business system, and establishes a local graph model according to the global initial graph model, the first user business feature vector and the second user attribute matrix issued by the central server;

[0008] The business system server samples a plurality of subgraphs in the local graph model according to the connection relationship between the first user and the second user, and trains local model parameters according to the subgraphs and the global initial graph model, and uploads the local model parameters to the central server;

[0009] The central server calculates global model parameters according to the local model parameters uploaded by the business system servers and the registered user numbers of the business systems, and distributes the global model parameters to the business system servers.

[0010] The business system server obtains a user matching result according to the currently stored global model parameters.

[0011] According to another aspect of the present application, a user intelligent matching device is provided, which is executed by a central server and business system servers, and comprises:

[0012] A local graph model establishing module is configured to generate a first user business feature vector and a second user attribute matrix in a business system by the business system server, and establish a local graph model according to a global initial graph model distributed by the central server, the first user business feature vector and the second user attribute matrix.

[0013] A local model parameter generating module is configured to sample a plurality of sub-graphs in the local graph model according to a connection relationship between a first user and a second user by the business system server, and train local model parameters according to the sub-graphs and the global initial graph model, and upload the local model parameters to the central server.

[0014] A global model parameter generating module is configured to calculate global model parameters according to the local model parameters uploaded by the business system servers and the registered user numbers of the business systems by the central server, and distribute the global model parameters to the business system servers.

[0015] A user matching module is configured to obtain a user matching result according to the currently stored global model parameters by the business system server.

[0016] According to another aspect of the present application, an electronic device is provided, which comprises:

[0017] at least one processor; and

[0018] a memory connected with the at least one processor; wherein

[0019] 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 to enable the at least one processor to execute the user intelligent matching method according to any one of the embodiments of the present application.

[0020] According to another aspect of the present application, a computer readable storage medium is provided, which stores computer instructions for enabling a processor to execute the user intelligent matching method according to any one of the embodiments of the present application.

[0021] The technical solution of this invention generates a first user business feature vector and a second user attribute matrix in the business system through a business system server. Based on the global initial graph model, the first user business feature vector, and the second user attribute matrix issued by the central server, a local graph model is established. Multiple subgraphs are sampled in this local graph model based on the connection relationship between the first and second users. Local model parameters are trained based on the subgraphs and the global initial graph model. These local model parameters are then uploaded to the central server. The central server calculates global model parameters based on the local model parameters uploaded by each business system server and the number of registered users in each business system. The global model parameters are then distributed to each business system server. User matching results are obtained based on the currently stored global model parameters. This method effectively integrates multi-source heterogeneous data for model training while ensuring data security in each business system, effectively improving the model's matching accuracy and adapting to dynamic business rule adjustments.

[0022] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart of a user intelligent matching method provided according to Embodiment 1 of the present invention;

[0025] Figure 2 This is a flowchart of another user intelligent matching method provided according to Embodiment 2 of the present invention;

[0026] Figure 3 This is a schematic diagram of the structure of a user intelligent matching device according to Embodiment 3 of the present invention;

[0027] Figure 4 This is a schematic diagram of the structure of an electronic device that implements the user intelligent matching method of this invention. Detailed Implementation

[0028] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0029] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0030] Example 1

[0031] Figure 1 This is a flowchart of a user intelligent matching method provided in Embodiment 1 of the present invention. This embodiment is applicable to situations where optimal model parameters are obtained by combining data from multiple business systems to accurately match users. This method can be executed by a user intelligent matching device, which can be implemented in hardware and / or software and is generally configured in a computer or processor with data processing capabilities. Figure 1 As shown, the method includes:

[0032] S110. Through the business system server, generate the first user business feature vector and the second user attribute matrix in the business system, and establish a local graph model based on the global initial graph model, the first user business feature vector and the second user attribute matrix issued by the central server.

[0033] Optionally, the central server can refer to the core server that coordinates collaborative modeling, parameter aggregation, and policy control across multiple business systems; the business system server can refer to the local server of each branch or business unit.

[0034] Optionally, in the user intelligent matching method proposed in this invention, a first user is matched with a second user who has business needs. The first user can provide consultation or guidance services to the second user to meet the second user's business requirements.

[0035] Optionally, the first user business feature vector is a multi-dimensional vector that quantifies the first user's business capabilities and attributes; the second user attribute matrix is ​​a standardized set of second user features.

[0036] Optionally, the global initial graph model is a pre-built graph neural network (GNN) basic framework on the central server, which includes initial model parameters and graph rules, but does not contain any original business data; the local graph model is a graph structure built by the business system server based on its own data, which includes the first user node, the second user node and the business relationship edge between them with bound features.

[0037] The generation of the first user business feature vector and the second user attribute matrix in the business system through the business system server may include:

[0038] Based on the historical business data maintained by the business system server, multiple business indicators of the first user and multi-dimensional user characteristics of the second user are extracted respectively.

[0039] The business system server generates business feature vectors for each first user based on multiple business metrics of the first user and the current dynamic weight coefficients of each business metric.

[0040] The business system server standardizes the multi-dimensional user features and generates a second user attribute matrix based on the standardized multi-dimensional user features.

[0041] Optionally, the first user's multiple business metrics may include, but are not limited to, business stability, compliance score, professional qualification level, online conversion rate, etc., and the second user's multi-dimensional user characteristics may include, but are not limited to, activity level, retention rate, number of business transactions, etc. The specific business metrics and multi-dimensional user characteristics can be determined according to the actual matching needs. This is only an example.

[0042] Optionally, the current dynamic weighting coefficients of each business indicator can be obtained from the central server or maintained independently in the business system.

[0043] The local graph model, established based on the global initial graph model, the first user business feature vector, and the second user attribute matrix issued by the central server, may include:

[0044] Based on the graph rules, first user business feature vector, and second user attribute matrix in the global initial graph model, the business system server establishes first user nodes and second user nodes in this map model, and establishes business relationship edges between the first user nodes and the second user nodes.

[0045] The business system server calculates the edge weights of the business relationship edges based on the historical business data of each group of first and second user nodes, and adds the edge weights to the local graph model.

[0046] Optionally, the graph rules may include node construction rules and edge construction rules. Node construction rules may include that the first user node must be bound to the first user's business feature vector and the second user node must be bound to the second user's attribute matrix. Edge construction rules may include the business relationship edge construction logic between users.

[0047] Optionally, based on the first user's business feature vector and the second user's attribute matrix, a corresponding first user node can be created for each first user in the local graph model, with its business feature vector as the node attribute. Similarly, a corresponding second user node can be created for each second user, with its user attribute matrix as the node attribute. Based on the actual matching situation of users in historical business data, business relationship edges can be established between users. The actual matching situation of users includes the matching situation between the first user and the second user, as well as the matching situation between the second users. There is also a business relationship between two second users with a matching relationship. Therefore, the business relationship edge connects the first user node and the second user node, and can also connect two different second user nodes.

[0048] Optionally, edge weight can be a numerical value that quantifies the strength of business relationships. It is calculated as the proportion of the number of historical business interactions between two users to the total number of interactions. For example, if user A and user B interact 10 times, of which 5 times result in business interactions, the edge weight is 5 / 10 = 0.5. The business system server assigns the calculated edge weight to the corresponding business relationship edge and adds it to the local graph model to complete the construction of the local graph model.

[0049] S120. Through the business system server, based on the connection relationship between the first user and the second user, multiple sub-graphs are sampled in this map model, and based on the sub-graphs and the global initial graph model, local model parameters are trained and uploaded to the central server.

[0050] Optionally, the connection between the first user and the second user refers to the established business relationship edge between the first user node and the second user node in the local graph model; the target first user node can refer to any selected first user node; the multi-level neighbor node can refer to the k-hop neighbor node of the target first user node. For example, k can be 3, where the 1-hop neighbor node is the second user node directly connected to the target first user node, the 2-hop neighbor node is another second user node directly connected to the 1-hop neighbor node, and the 3-hop neighbor node is another second user node directly connected to the 2-hop neighbor node. If the target first user node is node A, its 1-hop neighbors are nodes B and C, its 2-hop neighbors are nodes D (connected to B) and E (connected to C), and its 3-hop neighbors are nodes F (connected to D) and G (connected to E), then A, B, C, D, E, F, and G are extracted to form a subgraph. Multiple subgraphs are obtained by sampling each target first user node in this way.

[0051] Specifically, through the business system server, multiple sub-graphs are sampled in this map model based on the connection relationship between the first user and the second user. Based on these sub-graphs and the global initial graph model, local model parameters are trained, which may include:

[0052] Based on the connection relationship between the first user and the second user, the multi-level neighbor nodes of the target first user node are determined in this map model through the business system server, and the subgraph composed of the target first user node and its multi-level neighbor nodes is extracted.

[0053] The business system server updates the node features in each subgraph based on the initial features of each node and the edge weights between every two nodes.

[0054] In each training round, the target subgraph with updated features is input into the global initial graph model via the business system server. Based on the output of the global initial graph model and historical business data, the model loss for the current round is calculated. Based on the model loss for the current round and the pre-generated noise, the initial model parameters in the global initial graph model are updated for use in the next training round.

[0055] When the training termination conditions are met, the current model parameters are set as local model parameters by the business system server.

[0056] Optionally, the business system server can use the GraphSAGE algorithm to perform feature aggregation on nodes in each subgraph. Specifically, this includes: taking the target node in the subgraph as the core, performing a weighted average of the initial features of the neighboring nodes based on the edge weights between the target node and its neighboring nodes, and then fusing the weighted average result with the node's own initial features to obtain the updated node features.

[0057] Optionally, in each training round, the business system server inputs the feature-updated subgraph into the global initial graph model, and the model outputs the matching probability between the first user and the second user. Then, based on the business transaction tags in the historical business data, the cross-entropy loss function is used to calculate the model loss for the current round, where the tags of actual business transactions are marked as 1, and otherwise as 0. Based on the model loss for the current round, the initial model parameters are adjusted, and Laplacian noise is injected to ensure that differential privacy requirements are met and to prevent the original data from being obtained by reverse engineering.

[0058] Optionally, the business system server repeats the above training process. When the training termination condition is met, such as all subgraphs participating in training or the model loss value decreasing by less than the preset loss for multiple consecutive rounds, training is stopped, and the current model parameters are determined as the local model parameters.

[0059] S130. The central server calculates the global model parameters based on the local model parameters uploaded by each business system server and the number of registered users in each business system, and then distributes the global model parameters to each business system server.

[0060] The global model parameters are calculated by the central server based on the local model parameters uploaded by each business system server and the number of registered users in each business system. These parameters may include:

[0061] The parameter weights of each business system are determined by the number of registered users in each business system through the central server.

[0062] Using a central server and a federated averaging algorithm, the local model parameters uploaded by each business system server and the parameter weights of each business system are weighted and calculated to generate global model parameters.

[0063] Optionally, the central server can receive local model parameters from all business system servers and allocate weights according to the proportion of data volume of each business system. Specifically, it can calculate the total number of registered users in each business system, calculate the ratio of the number of registered users in the target business system to the total number of registered users, and use the ratio calculation result as the parameter weight of the target business system.

[0064] S140. Obtain user matching results through the business system server based on the currently stored global model parameters.

[0065] Before retrieving user matching results from the business system server based on currently stored global model parameters, the process may also include:

[0066] The business system server divides the second users in the business system into at least two groups, and uses the currently stored global model parameters to perform user matching for each group.

[0067] The matching accuracy of each group is obtained through the business system server based on the matching results of each group and historical business data, and the matching accuracy of each group is used to perform first-level verification.

[0068] The system server randomly generates interfering factors from user characteristics and performs user matching again using the currently stored global model parameters, and performs secondary verification based on the matching results.

[0069] Optionally, the second user can be divided into at least two groups based on the differences in the characteristics of the second user. For example, the user can be divided into a multi-business user group and a regular user group. The matching rules corresponding to the global model parameters are used for matching for different groups. Based on the actual matching results of each group and the manual matching results in historical business data, it is determined whether the matching accuracy of each group reaches the preset threshold and whether the maximum difference in the matching accuracy of each group is less than the preset first threshold. If both are satisfied, the first-level verification is passed.

[0070] Optionally, interfering factors can refer to simulated variables not included in the original user features, such as weather impact coefficients, business processing procedures, etc. After adding interfering factors to the features of the second user, the global model parameters can be called again for matching to calculate a new matching accuracy. The difference between the matching accuracy of each group without added interfering factors and the matching accuracy with added interfering factors is calculated. If the difference is less than the preset second threshold, the second-level verification is passed.

[0071] Optionally, the above verification method can effectively ensure the stability of the global model parameters. Only after the verification is passed can the user matching operation be performed using the verified global model parameters.

[0072] Optionally, the business system server can call the currently stored global model parameters to perform matching calculations between the first user and the second user to be matched, and obtain the user matching results.

[0073] The technical solution of this invention generates a first user business feature vector and a second user attribute matrix in the business system through a business system server. Based on the global initial graph model, the first user business feature vector, and the second user attribute matrix issued by the central server, a local graph model is established. Multiple subgraphs are sampled in this local graph model based on the connection relationship between the first and second users. Local model parameters are trained based on the subgraphs and the global initial graph model. These local model parameters are then uploaded to the central server. The central server calculates global model parameters based on the local model parameters uploaded by each business system server and the number of registered users in each business system. The global model parameters are then distributed to each business system server. User matching results are obtained based on the currently stored global model parameters. This method effectively integrates multi-source heterogeneous data for model training while ensuring data security in each business system, effectively improving the model's matching accuracy and adapting to dynamic business rule adjustments.

[0074] Example 2

[0075] Figure 2 This is a flowchart of a user intelligent matching method provided in Embodiment 2 of the present invention. This embodiment specifically describes the user intelligent matching method based on the above embodiments. Figure 2 As shown, the method includes:

[0076] S210. Using the business system server, extract multiple business indicators for the first user and multidimensional user characteristics for the second user based on the historical business data maintained by the business system.

[0077] S220. Through the business system server, generate business feature vectors for each first user based on multiple business indicators of the first user and the current dynamic weight coefficients of each business indicator.

[0078] S230. Through the business system server, the multi-dimensional user features are standardized, and a second user attribute matrix is ​​generated based on the standardized multi-dimensional user features.

[0079] S240. Through the business system server, based on the graph rules in the global initial graph model, the first user business feature vector, and the second user attribute matrix, establish the first user node and the second user node in this map model, and establish the business relationship edge between the first user node and the second user node.

[0080] S250. Through the business system server, calculate the edge weight of the business relationship edge based on the historical business data of each group of first user nodes and second user nodes, and add the edge weight to the local graph model.

[0081] S260. Through the business system server, based on the connection relationship between the first user and the second user, multiple sub-graphs are sampled in this map model, and based on the sub-graphs and the global initial graph model, local model parameters are trained and uploaded to the central server.

[0082] S270. The central server calculates the global model parameters based on the local model parameters uploaded by each business system server and the number of registered users in each business system, and then distributes the global model parameters to each business system server.

[0083] S280. Through the business system server, the second user in the business system is divided into at least two groups, and user matching is performed on each group using the currently stored global model parameters.

[0084] S290. Through the business system server, based on the matching results of each group and historical business data, obtain the matching accuracy of each group, and use the matching accuracy of each group to perform first-level verification.

[0085] S2100: Through the business system server, random interference factors are generated in the user characteristics, and user matching is performed again using the currently stored global model parameters. Secondary verification is performed based on the matching results.

[0086] S2110. Obtain user matching results through the business system server based on the currently stored global model parameters.

[0087] Furthermore, the user intelligent matching method may also include:

[0088] Whenever the business system server receives a policy correction rule from the central server, it corrects the currently stored global model parameters according to the policy correction rule.

[0089] Optionally, a central server can synchronize business information and data compliance status every natural day, formulate data security rules, generate policy correction rules based on data compliance status, and then distribute the data security rules and policy correction rules to the servers of each business system.

[0090] Optionally, when there is data interaction between the business system server and the central server, the data sender calls the currently stored data security rules to encrypt the data to be sent. The data security rules may include encryption algorithms and access token validity periods.

[0091] The technical solution of this invention generates a first user business feature vector and a second user attribute matrix in the business system through a business system server. Based on the global initial graph model, the first user business feature vector, and the second user attribute matrix issued by the central server, a local graph model is established. Multiple subgraphs are sampled in this local graph model based on the connection relationship between the first and second users. Local model parameters are trained based on the subgraphs and the global initial graph model. These local model parameters are then uploaded to the central server. The central server calculates global model parameters based on the local model parameters uploaded by each business system server and the number of registered users in each business system. The global model parameters are then distributed to each business system server. User matching results are obtained based on the currently stored global model parameters. This method effectively integrates multi-source heterogeneous data for model training while ensuring data security in each business system, effectively improving the model's matching accuracy and adapting to dynamic business rule adjustments.

[0092] Example 3

[0093] Figure 3 This is a schematic diagram of a user intelligent matching device provided in Embodiment 3 of the present invention. Figure 3 As shown, the device includes: a local graph model building module 310, a local model parameter generation module 320, a global model parameter generation module 330, and a user matching module 340.

[0094] The map model building module 310 is used to generate the first user business feature vector and the second user attribute matrix in the business system through the business system server, and to build the local map model based on the global initial map model, the first user business feature vector and the second user attribute matrix issued by the central server.

[0095] The local model parameter generation module 320 is used to sample multiple sub-graphs in the local map model through the business system server based on the connection relationship between the first user and the second user, and to train local model parameters based on the sub-graphs and the global initial graph model, and then upload the local model parameters to the central server.

[0096] The global model parameter generation module 330 is used to calculate the global model parameters through the central server based on the local model parameters uploaded by each business system server and the number of registered users of each business system, and then distribute the global model parameters to each business system server.

[0097] The user matching module 340 is used to obtain user matching results through the business system server based on the currently stored global model parameters.

[0098] The technical solution of this invention generates a first user business feature vector and a second user attribute matrix in the business system through a business system server. Based on the global initial graph model, the first user business feature vector, and the second user attribute matrix issued by the central server, a local graph model is established. Multiple subgraphs are sampled in this local graph model based on the connection relationship between the first and second users. Local model parameters are trained based on the subgraphs and the global initial graph model. These local model parameters are then uploaded to the central server. The central server calculates global model parameters based on the local model parameters uploaded by each business system server and the number of registered users in each business system. The global model parameters are then distributed to each business system server. User matching results are obtained based on the currently stored global model parameters. This method effectively integrates multi-source heterogeneous data for model training while ensuring data security in each business system, effectively improving the model's matching accuracy and adapting to dynamic business rule adjustments.

[0099] Based on the above embodiments, the local graph model building module 310 can be used for:

[0100] Based on the historical business data maintained by the business system server, multiple business indicators of the first user and multi-dimensional user characteristics of the second user are extracted respectively.

[0101] The business system server generates business feature vectors for each first user based on multiple business metrics of the first user and the current dynamic weight coefficients of each business metric.

[0102] The business system server standardizes the multi-dimensional user features and generates a second user attribute matrix based on the standardized multi-dimensional user features.

[0103] Based on the above embodiments, the local graph model building module 310 can be further used for:

[0104] Based on the graph rules, first user business feature vector, and second user attribute matrix in the global initial graph model, the business system server establishes first user nodes and second user nodes in this map model, and establishes business relationship edges between the first user nodes and the second user nodes.

[0105] The business system server calculates the edge weights of the business relationship edges based on the historical business data of each group of first and second user nodes, and adds the edge weights to the local graph model.

[0106] Based on the above embodiments, the local model parameter generation module 320 can be specifically used for:

[0107] Based on the connection relationship between the first user and the second user, the multi-level neighbor nodes of the target first user node are determined in this map model through the business system server, and the subgraph composed of the target first user node and its multi-level neighbor nodes is extracted.

[0108] The business system server updates the node features in each subgraph based on the initial features of each node and the edge weights between every two nodes.

[0109] In each training round, the target subgraph with updated features is input into the global initial graph model via the business system server. Based on the output of the global initial graph model and historical business data, the model loss for the current round is calculated. Based on the model loss for the current round and the pre-generated noise, the initial model parameters in the global initial graph model are updated for use in the next training round.

[0110] When the training termination conditions are met, the current model parameters are set as local model parameters by the business system server.

[0111] Based on the above embodiments, the global model parameter generation module 330 can be specifically used for:

[0112] The parameter weights of each business system are determined by the number of registered users in each business system through the central server.

[0113] Using a central server and a federated averaging algorithm, the local model parameters uploaded by each business system server and the parameter weights of each business system are weighted and calculated to generate global model parameters.

[0114] Based on the above embodiments, a model verification module may also be included, for:

[0115] The business system server divides the second users in the business system into at least two groups, and uses the currently stored global model parameters to perform user matching for each group.

[0116] The matching accuracy of each group is obtained through the business system server based on the matching results of each group and historical business data, and the matching accuracy of each group is used to perform first-level verification.

[0117] The system server randomly generates interfering factors from user characteristics and performs user matching again using the currently stored global model parameters, and performs secondary verification based on the matching results.

[0118] Based on the above embodiments, a model correction module may also be included, used for:

[0119] Whenever the business system server receives a policy correction rule from the central server, it corrects the currently stored global model parameters according to the policy correction rule.

[0120] The user intelligent matching device provided in the embodiments of the present invention can execute the user intelligent matching method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the method execution.

[0121] Example 4

[0122] Figure 4 A schematic diagram of an electronic device 10, which can be used to implement embodiments of the present invention, is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0123] like Figure 4 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0124] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0125] Processor 11 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, central processing unit (CPU), graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as the user intelligent matching method described in the embodiments of the present invention. That is:

[0126] The business system server generates the first user business feature vector and the second user attribute matrix in the business system, and establishes a local graph model based on the global initial graph model, the first user business feature vector and the second user attribute matrix issued by the central server.

[0127] Based on the connection relationship between the first user and the second user, the business system server samples multiple sub-graphs in the map model and trains local model parameters based on the sub-graphs and the global initial graph model. The local model parameters are then uploaded to the central server.

[0128] The central server calculates the global model parameters based on the local model parameters uploaded by each business system server and the number of registered users in each business system, and then distributes the global model parameters to each business system server.

[0129] The system server retrieves user matching results based on the currently stored global model parameters.

[0130] In some embodiments, the user intelligent matching method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the user intelligent matching method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to execute the user intelligent matching method by any other suitable means (e.g., by means of firmware).

[0131] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0132] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0133] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0134] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0135] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0136] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0137] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0138] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A user intelligent matching method, executed by a central server in cooperation with a plurality of service system servers, characterized by, The method comprises the following steps: generating, by the business system server, a first user business feature vector and a second user attribute matrix in the business system, and establishing a local graph model according to a global initial graph model issued by the central server, the first user business feature vector and the second user attribute matrix; sampling, by the business system server, a plurality of sub-graphs in the local graph model according to the connection relationship between the first user and the second user, and training local model parameters according to the sub-graphs and the global initial graph model, and uploading the local model parameters to the central server; calculating, by the central server, global model parameters according to the local model parameters uploaded by each business system server and the number of registered users of each business system, and issuing the global model parameters to each business system server; obtaining, by the business system server, a user matching result according to the currently stored global model parameters.

2. The method of claim 1, wherein, The method comprises the following steps: extracting, by the business system server, a plurality of business indicators of the first user and a plurality of user features of the second user from historical business data maintained by the business system; generating, by the business system server, a first user business feature vector according to the plurality of business indicators of the first user and the current dynamic weight coefficient of each business indicator; standardizing, by the business system server, the plurality of user features, and generating a second user attribute matrix according to the standardized plurality of user features.

3. The method of claim 1, wherein, The method comprises the following steps: establishing, by the business system server, a first user node and a second user node in the local graph model according to the graph rules in the global initial graph model, the first user business feature vector and the second user attribute matrix, and establishing a business relationship edge between the first user node and the second user node; calculating, by the business system server, an edge weight of the business relationship edge according to historical business data of each group of first user nodes and second user nodes, and adding the edge weight to the local graph model.

4. The method of claim 1, wherein, The method comprises the following steps: determining, by the business system server, a plurality of level neighbor nodes of a target first user node in the local graph model according to the connection relationship between the first user and the second user, and extracting a sub-graph composed of the target first user node and the plurality of level neighbor nodes; updating, by the business system server, the node features in each sub-graph according to the initial features of the nodes in the sub-graph and the edge weight between each two nodes; The business system server inputs the target subgraph with the updated features into the global initial graph model in each round of training, calculates the model loss of the current round according to the output result of the global initial graph model and historical business data, and updates the initial model parameters in the global initial graph model according to the model loss of the current round and the pre-generated noise for use in the next round of training. The business system server determines the current model parameters as the local model parameters when it is determined that the training end condition is met.

5. The method of claim 1, wherein, The central server calculates the global model parameters according to the local model parameters uploaded by the business system servers and the number of registered users of each business system, including: The central server determines the parameter weights of each business system according to the number of registered users of each business system. The central server uses a federated averaging algorithm to perform weighted calculation on the local model parameters uploaded by the business system servers and the parameter weights of each business system to generate the global model parameters.

6. The method of claim 1, wherein, Before the business system server obtains the user matching result according to the currently stored global model parameters, it further includes: The business system server divides the second users in the business system into at least two groups, and performs user matching on each group using the currently stored global model parameters. The business system server obtains the matching accuracy of each group according to the matching result of each group and historical business data, and performs primary verification using the matching accuracy of each group. The business system server randomly generates interference factors in the user features, and performs user matching again using the currently stored global model parameters, and performs secondary verification according to the matching result.

7. The method of claim 1, wherein, Further including: The business system server modifies the currently stored global model parameters according to the policy correction rule issued by the central server.

8. A user intelligent matching apparatus, executed by a central server in cooperation with each service system server, characterized by, Including: The local graph model establishment module is configured to generate the first user business feature vector and the second user attribute matrix in the business system by the business system server, and establish a local graph model according to the global initial graph model issued by the central server, the first user business feature vector, and the second user attribute matrix; The local model parameter generation module is configured to sample a plurality of subgraphs in the local graph model according to the connection relationship between the first user and the second user by the business system server, and train the local model parameters according to the subgraphs and the global initial graph model, and upload the local model parameters to the central server; The global model parameter generation module is configured to calculate the global model parameters according to the local model parameters uploaded by the business system servers and the number of registered users of each business system by the central server, and issue the global model parameters to each business system server; The user matching module is configured to obtain the user matching result according to the currently stored global model parameters by the business system server.

9. An electronic device, comprising: The electronic device includes: at least one processor; and a memory connected to the at least one processor in communication; 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 to enable the at least one processor to perform the user intelligent matching method in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer instructions for enabling the processor to implement the user intelligent matching method in any one of claims 1-7 when executed.

Images