Training method and device of suggestion generation model, model application method and device and medium

Abstract

The invention provides a training method of a suggestion generation model, which can be applied to the technical field of artificial intelligence. The training method of the suggestion generation model comprises the following steps: receiving a plurality of pieces of service feature data which are sent by a plurality of participants to a central server and are used for representing service interaction among the plurality of participants; clustering processing is performed on the multiple pieces of service feature data used for representing service interaction among the multiple participants by using a clustering algorithm to obtain multiple target service data sets; respectively inputting the plurality of target business data sets into the corresponding initial class cluster models to train each initial class cluster model to obtain a plurality of target class cluster models; and according to the model parameters of each target class cluster model, training an untrained suggestion generation model by using a federated learning algorithm to obtain a suggestion generation model for generating business processing suggestions for the participants. The invention further provides a model application method and device and a medium.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence, and more specifically to the training of a suggestion generation model, and more specifically to a training method of the suggestion generation model, an application method, a device and a medium of the model. Background Art

[0002] Federated learning is a distributed machine learning method that aims to uniformly train machine learning models based on business feature data provided by multiple participants. In related technologies, corresponding learning models are constructed based on the business feature data provided by each participant. By training each learning model, a recommendation generation model for the business scenario is obtained.

[0003] In the process of realizing the concept disclosed in the present invention, the inventors found that there are at least the following problems in the related technology: since the business feature data between multiple participants are independent of each other and the data structure of the business feature data is inconsistent, the correlation between the business feature data is not paid attention to during the model training process, resulting in the poor effect of the trained suggestion generation model on business processing. Summary of the invention

[0004] In view of the above problems, the present disclosure provides a training method for a suggestion generation model, an application method, a device and a medium of the model.

[0005] According to the first aspect of the present disclosure, a training method for a suggestion generation model is provided, comprising: receiving a plurality of business feature data sent by a plurality of participants to a central server for characterizing business interactions between the plurality of the above-mentioned participants; clustering the plurality of business feature data for characterizing business interactions between the plurality of the above-mentioned participants using a clustering algorithm to obtain a plurality of target business data sets; inputting the plurality of the above-mentioned target business data sets into corresponding initial cluster models respectively to train each of the above-mentioned initial cluster models to obtain a plurality of target cluster models; and training an untrained suggestion generation model using a federated learning algorithm according to model parameters of each of the above-mentioned target cluster models to obtain a suggestion generation model for generating business processing suggestions for the above-mentioned participants.

[0006] According to an embodiment of the present disclosure, the clustering algorithm is used to cluster the multiple business feature data used to characterize the business interaction between the multiple participants to obtain multiple target business data sets, including: analyzing the multiple business feature data to determine a data center point, wherein the data center point is the center coordinate point of the multiple business feature data in the data distribution space; dividing the multiple business feature data into multiple sub-data sets according to the distance between the data center point and each of the business feature data in the data distribution space, wherein the number of the business feature data in each of the sub-data sets is less than a preset number; determining the center of the minimum circumscribed circle of each of the sub-data sets in the data distribution space, so as to merge the multiple sub-data sets according to the distance between the multiple center points, to obtain multiple target business data sets, wherein the center distance between the multiple target business data sets is greater than the neighborhood radius.

[0007] According to an embodiment of the present disclosure, the above-mentioned multiple business feature data are divided into multiple sub-data sets according to the distance between the above-mentioned data center point and each of the above-mentioned business feature data in the above-mentioned data distribution space, including: based on the data coordinate point of each of the above-mentioned business feature data in the above-mentioned data distribution space, the data coordinate point farthest from the above-mentioned data center point is determined as a first-category cluster point; from the multiple data coordinate points, the data coordinate point farthest from the above-mentioned first-category cluster point is determined as a second-category cluster point: the vertical point from the above-mentioned data center point to the line connecting the above-mentioned first-category cluster point and the above-mentioned second-category cluster point is calculated; from the multiple data coordinate points, the data coordinate point farthest from the above-mentioned vertical point is determined as a third-category cluster point: the distances between the remaining data coordinate points and the above-mentioned first-category cluster point, the above-mentioned second-category cluster point and the above-mentioned third-category cluster point are calculated respectively, so as to divide the above-mentioned business feature data into multiple sub-data sets according to the principle of shortest distance.

[0008] According to an embodiment of the present disclosure, the business feature data is divided into the plurality of sub-datasets according to the principle of the shortest distance, including: analyzing the plurality of distances to divide the business feature data into sub-datasets corresponding to the first type of cluster points, sub-datasets corresponding to the second type of cluster points, and sub-datasets corresponding to the third type of cluster points according to the principle of the shortest distance; for each of the sub-datasets, when it is determined that the amount of the business feature data in the sub-dataset is greater than or equal to the preset amount, repeatedly performing the following steps to recursively divide the sub-datasets until the amount of the business feature data in the sub-datasets is less than the preset amount: determining new first type of cluster points, new second type of cluster points, and new third type of cluster points in the sub-datasets; and dividing the business feature data in the sub-datasets according to the distances between the remaining data coordinate points in the sub-datasets and the new first type of cluster points, the new second type of cluster points, and the new third type of cluster points, respectively, using the principle of the shortest distance to obtain a plurality of new sub-datasets.

[0009] According to an embodiment of the present disclosure, the above-mentioned multiple sub-data sets are merged according to the distances between the above-mentioned multiple circle centers to obtain multiple target business data sets, including: determining the coordinate distance between each of the above-mentioned business feature data and the remaining business feature data in the above-mentioned data distribution space; averaging the multiple above-mentioned coordinate distances to obtain the above-mentioned neighborhood radius; and merging the above-mentioned sub-data sets whose distances between the above-mentioned circle centers are smaller than the above-mentioned neighborhood radius in pairs to obtain the above-mentioned multiple target business data sets.

[0010] According to an embodiment of the present disclosure, the above-mentioned sub-datasets whose distances between the centers are less than the neighborhood radius are merged in pairs to obtain the above-mentioned multiple target business data sets, including: calculating the distance between the center of each of the above-mentioned sub-datasets and the center of the remaining sub-datasets; for each pair of the above-mentioned sub-datasets, when it is determined that the distance between the centers is less than the above-mentioned neighborhood radius, merging the two above-mentioned sub-datasets to obtain a new sub-dataset; determining the center of the minimum circumscribed circle corresponding to the above-mentioned new sub-dataset, and calculating the distance between the center of the above-mentioned new sub-dataset and the center of the remaining sub-datasets; when it is determined that there is a sub-dataset in the remaining sub-datasets whose distance to the center of the above-mentioned new sub-dataset is less than the above-mentioned neighborhood radius, iteratively merging the sub-datasets whose distances to the center of the above-mentioned new sub-dataset are less than the above-mentioned neighborhood radius based on the center of the above-mentioned new sub-dataset; when it is determined that the distances between the center of the above-mentioned new sub-dataset and the center of the remaining sub-datasets are all greater than or equal to the above-mentioned neighborhood radius, determining the multiple sub-datasets obtained after merging as the target business data sets.

[0011] According to an embodiment of the present disclosure, the above-mentioned data center point is determined by analyzing the multiple business characteristic data, including: determining the data coordinate point of each of the above-mentioned business characteristic data in the above-mentioned data distribution space; averaging the multiple data coordinate points to obtain the above-mentioned data center point.

[0012] The second aspect of the present disclosure provides a model application method, including: obtaining business feature data sent by a participant; inputting the above business feature data into the above suggestion generation model for analysis, and outputting the analysis results, wherein the above analysis results are used to characterize the business processing suggestions corresponding to the above business feature data, and the above participants can process the business according to the above business processing suggestions.

[0013] The third aspect of the present disclosure provides a training device for a suggestion generation model, comprising: a data receiving module for receiving a plurality of business feature data sent by a plurality of participants to a central server for characterizing business interactions between the plurality of the above-mentioned participants; a data clustering module for clustering the plurality of business feature data for characterizing business interactions between the plurality of the above-mentioned participants using a clustering algorithm to obtain a plurality of target business data sets; a model training module for inputting the plurality of the above-mentioned target business data sets into corresponding initial cluster models respectively to train each of the above-mentioned initial cluster models to obtain a plurality of target cluster models; a model building module for training an untrained suggestion generation model using a federated learning algorithm according to the model parameters of each of the above-mentioned target cluster models to obtain a suggestion generation model for generating business processing suggestions for the above-mentioned participants.

[0014] The fourth aspect of the present disclosure provides a model application device, including: an acquisition module, used to obtain business feature data sent by a participant; an analysis module, used to input the above-mentioned business feature data into the above-mentioned suggestion generation model for analysis, and output analysis results, wherein the above-mentioned analysis results are used to characterize business processing suggestions corresponding to the above-mentioned business feature data, and the above-mentioned participant can process the business according to the above-mentioned business processing suggestions.

[0015] A fifth aspect of the present disclosure provides an electronic device, comprising: one or more processors; and a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the above method.

[0016] The sixth aspect of the present disclosure further provides a computer-readable storage medium having a computer program or instructions stored thereon, which implements the steps of the above method when the above computer program or instructions are executed by a processor.

[0017] The seventh aspect of the present disclosure further provides a computer program product, including a computer program or instructions, which implement the steps of the above method when the above computer program or instructions are executed by a processor.

[0018] According to the embodiments of the present disclosure, a clustering algorithm is used to cluster multiple business feature data representing business interactions between multiple participants, and the mutually independent business feature data between multiple participants can be classified according to the internal associations and data structures between the business feature data. The multiple target cluster models obtained by training the initial cluster model using the target business data set can learn the internal associations between business feature data of various data structures. Therefore, according to the model parameters of each target cluster model, the suggestion generation model constructed using the federated learning algorithm can more accurately process the business feature data of each participant, thereby meeting business needs to the greatest extent. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The above contents and other objects, features and advantages of the present disclosure will become more apparent through the following description of the embodiments of the present disclosure with reference to the accompanying drawings, in which:

[0020] Figure 1 A diagram schematically illustrates an application scenario of a training method for a suggestion generation model, an application method of the model, a device, and a medium according to an embodiment of the present disclosure;

[0021] Figure 2 A flowchart of a method for training a suggestion generation model according to an embodiment of the present disclosure is schematically shown;

[0022] Figure 3 A schematic diagram of a process for processing service feature data according to an embodiment of the present disclosure is schematically shown;

[0023] Figure 4 A schematic diagram showing the relationship between the first type of cluster points, the second type of cluster points and the third type of cluster points according to an embodiment of the present disclosure is shown;

[0024] Figure 5 A flowchart of a method for dividing a sub-data set according to an embodiment of the present disclosure is schematically shown;

[0025] Figure 6 A flowchart of a method for determining a target business data set according to an embodiment of the present disclosure is schematically shown;

[0026] Figure 7 A schematic diagram of a suggestion generation model construction process according to an embodiment of the present disclosure is schematically shown;

[0027] Figure 8 A flowchart schematically shows a method for applying a model according to an embodiment of the present disclosure;

[0028] Fig. 9 The structural block diagram of the training device of the suggestion generation model according to the embodiment of the present disclosure is schematically shown;

[0029] Fig.10 A structural block diagram of a model application device according to an embodiment of the present disclosure is schematically shown; and

[0030] Fig.11 A block diagram of an electronic device suitable for implementing a training method for a suggestion generation model and / or an application method of the model according to an embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0031] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0032] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0033] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0034] When using expressions such as "at least one of A, B, and C, etc.", they should generally be interpreted according to the meaning of the expression commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0035] In the technical solution of the present disclosure, the user information (including but not limited to user personal information, user image information, user device information, such as location information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved are all information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data comply with relevant laws, regulations and standards, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0036] In the scenario of using personal information for automated decision-making, the methods, devices, and systems provided by the embodiments of the present disclosure provide users with corresponding operation portals for users to choose to agree or reject the automated decision-making results; if the user chooses to reject, the expert decision-making process will be entered. The expression "automated decision-making" here refers to the activity of automatically analyzing and evaluating a person's behavioral habits, interests and hobbies, or economic, health, credit status, etc. through computer programs, and making decisions. The expression "expert decision-making" here refers to the activity of making decisions by people who specialize in a certain field, have specialized experience, knowledge and skills, and have reached a certain level of professionalism.

[0037] Federated learning technology is widely used in various business application scenarios. In business application scenarios, federated learning models are used as business models to provide various business prediction services, such as classification prediction, business risk prediction, etc. In many cases, if federated learning does not use clustering algorithms when modeling, the data of each party exists independently, and the structure and rules between the data of different institutions cannot be discovered. Modeling can only be achieved when the data fields are known. If each party only provides data but not field information, the constructed federated learning model will be even less accurate and even less able to meet the needs.

[0038] The application of clustering algorithms in the construction of federated learning models is to target data from different parties with different characteristics, different distributions, and inconsistent data samples. DBSCAN (Density-Based Spatial Clustering of Applications with Noise) is a density-based spatial clustering algorithm. The DBSCAN algorithm divides areas with sufficient density into clusters, and finds clusters of any shape in a spatial database with noise, defining a cluster as the largest set of density-connected points.

[0039] The DBSCAN clustering algorithm has the following characteristics: DBSCAN is based on density clustering, which can find all dense areas of sample points and treat these dense areas as clusters one by one; the DBSCAN clustering algorithm mainly includes two algorithm parameters: neighborhood radius R and the minimum number of points (MinPoints). When the number of points within the neighborhood radius R is greater than the minimum number of points MinPoints, it is dense; the DBSCAN clustering algorithm contains three different types of points, namely core points, boundary points and noise points. Among them, the points whose number of sample points within the neighborhood radius R is greater than or equal to minpoints are called core points, points that are not core points but in the neighborhood of a core point are called boundary points, and those that are neither core points nor boundary points are noise points.

[0040] At present, the calculation process of the existing DBSCAN clustering algorithm mainly includes: 1. Find core points to form small clusters. By scanning all sample points, the number of points within the radius R of all sample points is obtained. When the number of points within the radius R of a certain sample point is greater than or equal to MinPoints, it is included in the core point list and the points directly reached by its density form the corresponding small cluster. 2. Merge small clusters to obtain large clusters. For each small cluster, check whether the point in it is a core point. If so, merge the small cluster corresponding to the point with the current small cluster to obtain a large cluster. 3. Repeat operations 1 and 2, and perform the same merging operation on the remaining small clusters until each point in the current small cluster is either not in the core point list, or the points directly reached by its density are already in the small cluster. The small cluster is upgraded to a large cluster, thereby obtaining the final clustering result.

[0041] The embodiments of the present disclosure provide a training method for a suggestion generation model, a model application method, a device and a medium. The training method for the suggestion generation model includes: receiving multiple business feature data sent by multiple participants to a central server for characterizing business interactions between multiple participants; clustering the multiple business feature data for characterizing business interactions between multiple participants using a clustering algorithm to obtain multiple target business data sets; inputting the multiple target business data sets into corresponding initial cluster models respectively to train each initial cluster model to obtain multiple target cluster models; and training an untrained suggestion generation model using a federated learning algorithm according to the model parameters of each target cluster model to obtain a suggestion generation model for generating business processing suggestions for the participants.

[0042] Figure 1 The application scenario diagram schematically shows the training method of the suggestion generation model, the application method of the model, the device and the medium according to the embodiment of the present disclosure.

[0043] like Figure 1As shown, the application scenario 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104 and a server 105. The network 104 is used to provide a medium for a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0044] The user can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications can be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only for example).

[0045] The first terminal device 101, the second terminal device 102, and the third terminal device 103 may be various electronic devices having display screens and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like.

[0046] The server 105 may be a server that provides various services, such as a background management server (only as an example) that provides support for websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103. The background management server may analyze and process the received data such as user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal device.

[0047] It should be noted that the training method of the suggestion generation model provided in the embodiment of the present disclosure can generally be executed by the server 105. Accordingly, the training device of the suggestion generation model provided in the embodiment of the present disclosure can generally be set in the server 105. The training method of the suggestion generation model provided in the embodiment of the present disclosure can also be executed by a server or a server cluster that is different from the server 105 and can communicate with the first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105. Accordingly, the training device of the suggestion generation model provided in the embodiment of the present disclosure can also be set in a server or a server cluster that is different from the server 105 and can communicate with the first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105.

[0048] It should be understood that Figure 1The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0049] The following will be based on Figure 1 The scene described by Figure 2~Figure 7 The training method of the suggestion generation model of the disclosed embodiment is described in detail.

[0050] Figure 2 The flowchart of the training method of the suggestion generation model according to the embodiment of the present disclosure is schematically shown.

[0051] like Figure 2 As shown, the training method of the suggestion generation model of this embodiment includes operations S210 to S240.

[0052] In operation S210, a plurality of business characteristic data sent by a plurality of participants to a central server and used to characterize business interactions between the plurality of participants are received.

[0053] In operation S220, a clustering algorithm is used to perform clustering processing on a plurality of business feature data used to characterize business interactions between a plurality of participants, to obtain a plurality of target business data sets.

[0054] In operation S230, a plurality of target business data sets are respectively input into corresponding initial cluster models to train each initial cluster model to obtain a plurality of target cluster models.

[0055] In operation S240, an untrained suggestion generation model is trained using a federated learning algorithm according to the model parameters of each target cluster model to obtain a suggestion generation model for generating business processing suggestions for the participants.

[0056] According to an embodiment of the present disclosure, before analyzing multiple business feature data, the consent or authorization of multiple participants can be obtained. For example, before operation S210, a request to obtain business feature data is sent to multiple participants. When multiple participants agree or authorize to obtain business feature data, operation S210 is performed. Due to data inconsistency, data heterogeneity and other problems in the multi-party joint modeling process, directly modeling the data will result in poor model effect and fail to meet business needs.

[0057] Figure 3 The following schematically shows a process diagram of processing service feature data according to an embodiment of the present disclosure. Figure 3As shown, before multiple parties jointly conduct the training of the suggestion generation model for federated learning, data clustering is performed through the central server 302, the data is divided into different clusters, and a model is constructed for each cluster to continuously optimize the model parameters, and finally the cluster model is summarized into the suggestion generation model to realize the construction of the suggestion generation model. The data source 303 of the federated learning task can be different devices or different industries, for example, the banking industry, the operator, and the e-commerce industry, and the business involved can be set according to actual needs, for example, commodity purchase. The types of business feature data can include structured data and unstructured data. In order to meet the security of business feature data not leaving the domain, in the process of business feature data union, the multi-party data is first homomorphically encrypted, and the encrypted data is then transmitted to the intermediate server for clustering operation.

[0058] According to an embodiment of the present disclosure, a variety of clustering algorithms may be used when clustering multiple business feature data representing business interactions between multiple parties, for example, the K-means algorithm, the K-center point algorithm or the DBSCAN algorithm, or a combination of multiple clustering algorithms.

[0059] According to an embodiment of the present disclosure, in the process of combining cluster models to determine a suggestion generation model, the parameters of the target cluster model can be divided into different clusters by clustering, and different clusters can be used for multiple rounds of iterative learning. Then, the target cluster model parameters selected in each round are weighted averaged according to the sample weights to obtain the suggestion generation model.

[0060] According to the embodiments of the present disclosure, a clustering algorithm is used to cluster multiple business feature data representing business interactions between multiple participants, and the mutually independent business feature data between multiple participants can be classified according to the internal associations and data structures between the business feature data. The multiple target cluster models obtained by training the initial cluster model using the target business data set can learn the internal associations between business feature data of various data structures. Therefore, according to the model parameters of each target cluster model, the suggestion generation model constructed using the federated learning algorithm can more accurately process the business feature data of each participant, thereby meeting business needs to the greatest extent.

[0061] According to an embodiment of the present disclosure, a clustering algorithm is used to perform clustering processing on multiple business feature data used to characterize business interactions between multiple participants to obtain multiple target business data sets, including: analyzing multiple business feature data to determine a data center point, wherein the data center point is the central coordinate point of the multiple business feature data in the data distribution space; dividing the multiple business feature data into multiple sub-data sets according to the distance between the data center point and each business feature data in the data distribution space, wherein the number of business feature data in each sub-data set is less than a preset number; determining the center of the minimum circumscribed circle of each sub-data set in the data distribution space, so as to merge the multiple sub-data sets according to the distance between the multiple center points, to obtain multiple target business data sets, wherein the center distance between the multiple target business data sets is greater than the neighborhood radius.

[0062] According to an embodiment of the present disclosure, the data center point can be determined by first determining the coordinates of the two points of the business feature data that are farthest apart in the data distribution space, and taking the midpoint of the coordinates of the two farthest points as the coordinates of the data center point.

[0063] According to an embodiment of the present disclosure, when the business feature data is divided into multiple sub-data sets, the division can be performed based on the distance between the data center point in the data distribution space and each business feature data, and the distance calculation method includes Euclidean distance, Chebyshev distance, Minkowski distance, standardized Euclidean distance, etc.

[0064] According to an embodiment of the present disclosure, by dividing business feature data into sub-data sets, and dividing business feature data with inconsistent data structures into multiple sub-data sets with the same data structure, the associations and regularities between the various business feature data can be better utilized in subsequent model training.

[0065] According to the embodiments of the present disclosure, Figure 3 As shown, when merging the sub-datasets 301 in the data distribution space, the data types, data sources or data structures of the business feature data in the sub-datasets 301 can be used, or the distance between the sub-datasets 301 can be used, or the similarity between the sub-datasets 301 can be used. By merging the sub-datasets 301, the connections and rules between the business feature data in the target business data set can be increased, so that the trained suggestion generation model can learn the connections and rules between the business feature data, so that more accurate analysis results can be obtained when the suggestion generation model is used for data analysis.

[0066] According to the embodiments of the present disclosure, by first dividing the business feature data into multiple sub-datasets and then merging the multiple sub-datasets to obtain the target business data set, in the face of uneven data distribution, the data clusters can be accurately identified by using a step-by-step division method, effectively reducing the occurrence of clustering bias. In the subsequent steps, the target business data set is used to train the initial cluster model, so that the training suggestion generation model is more accurate, and in the process of using the suggestion generation model to perform business processing, it can meet business needs to the greatest extent and improve business processing efficiency.

[0067] Figure 4 The diagram schematically shows the relationship between the first type of cluster points, the second type of cluster points and the third type of cluster points according to an embodiment of the present disclosure.

[0068] According to an embodiment of the present disclosure, multiple business feature data are divided into multiple sub-data sets according to the distance between the data center point and each business feature data in the data distribution space, including: based on the data coordinate point of each business feature data in the data distribution space, the data coordinate point farthest from the data center point is determined as the first type of cluster point; from multiple data coordinate points, the data coordinate point farthest from the first type of cluster point is determined as the second type of cluster point; the vertical point from the data center point to the line connecting the first type of cluster point and the second type of cluster point is calculated; from multiple data coordinate points, the data coordinate point farthest from the vertical point is determined as the third type of cluster point. The distances between the remaining data coordinate points and the first type of cluster point, the second type of cluster point and the third type of cluster point are calculated respectively, so as to divide the business feature data into multiple sub-data sets according to the principle of the shortest distance.

[0069] According to an embodiment of the present disclosure, the division of multiple service feature data into multiple sub-data sets may be based on the idea of ​​Ball-Tree. Figure 4 As shown in the figure, the data center point is point O. We can first calculate the point farthest from the data center point O as point A, which is the first type of cluster point. Calculate the point farthest from point A as point B, which is the second type of cluster point. Determine the line AB connecting points A and B, calculate the perpendicular point O' from point O to the line connecting AB, calculate the point farthest from point O' as point C, which is the third type of cluster point.

[0070] According to the embodiments of the present disclosure, in addition to the business feature data represented by point A, point B, and point C, the distances between other business feature data in the data distribution space and point A, point B, and point C are calculated respectively, and according to the principle of closest distance, the other business feature data in the data distribution space are divided into a first cluster, a second cluster, and a third cluster; wherein the first cluster includes the business feature data corresponding to point A and the business feature data closest to point A; the second cluster includes the business feature data corresponding to point B and the business feature data closest to point B; the third cluster includes the business feature data corresponding to point C and the business feature data closest to point C.

[0071] According to the embodiments of the present disclosure, when using the ball tree to divide the sub-datasets, each sub-dataset is a ball, which is a circle on a two-dimensional plane. When merging the sub-datasets, only the distance between the centers of the circles needs to be calculated, which improves the efficiency of subsequent calculations. Since the ball tree algorithm has low time complexity, it can significantly improve the calculation speed when dividing the sub-datasets, and can also adjust the parameters according to actual needs. The parameters are the number of business feature data. By dividing the business feature data into sub-datasets, the associations and rules between the business feature data can be better discovered, providing strong support for the analysis of the suggestion generation model.

[0072] Figure 5 A flowchart of a method for dividing a sub-data set according to an embodiment of the present disclosure is schematically shown. Figure 5 As shown, the sub-data set partitioning method of this embodiment includes operations S510 to S550.

[0073] In operation S510, first-category cluster points, new second-category cluster points, and new third-category cluster points are determined in the sub-dataset.

[0074] In operation S520, the distances between the remaining data coordinate points in the sub-data set and the first type of cluster points, the second type of cluster points, and the third type of cluster points are calculated.

[0075] In operation S530, the business characteristic data in the sub-datasets are divided according to the principle of closest distance to obtain a plurality of new sub-datasets.

[0076] In operation S540, determine whether the amount of business feature data in the sub-data set is greater than or equal to the preset amount; when the amount of business feature data in the sub-data set is greater than or equal to the preset amount, execute operation S510; when the amount of business feature data in the sub-data set is less than the preset amount, execute operation S550.

[0077] In operation S550, a plurality of sub-data sets are determined.

[0078] According to an embodiment of the present disclosure, the idea of ​​the Ball-Tree algorithm is applied, and each data point can be regarded as a sphere, each sphere represents a data category, and multiple cluster points are determined from multiple data points in the data set. The data points are divided by comparing the distances between the spheres, so that the data points represented by all business feature data are divided into multiple clusters, and the above steps are repeated to continue dividing each cluster.

[0079] According to the embodiment of the present disclosure, the first cluster, the second cluster and the third cluster are continuously recursively divided according to the example method to obtain sub-data sets in the first cluster, the second cluster and the third cluster, until the number of business feature data included in the sub-data sets is less than the preset number, the division of the business feature data is stopped to obtain multiple sub-data sets. The number of cluster points can be determined according to actual needs.

[0080] According to the embodiments of the present disclosure, by applying the idea of ​​the Ball-Tree algorithm, the business feature data is divided into multiple sub-data sets according to the internal connections between the data structure and the business feature data, so that it is convenient to apply the federated learning method in the subsequent steps to learn the connections between the business feature data in each sub-data set, thereby overcoming the defect of insufficient accuracy of the suggestion generation model caused by problems such as inconsistent business feature data structure and data heterogeneity.

[0081] According to an embodiment of the present disclosure, multiple sub-data sets are merged according to the distances between multiple circle centers to obtain multiple target business data sets, including: determining the coordinate distance between each business feature data and the remaining business feature data in the data distribution space. The multiple coordinate distances are averaged to obtain a neighborhood radius. Sub-data sets whose distances between the circle centers are less than the neighborhood radius are merged in pairs to obtain multiple target business data sets.

[0082] According to an embodiment of the present disclosure, the center of the data distribution space can be determined by determining the circumscribed circle of the data distribution space, and the coordinate distance between the remaining business feature data and the center of the circle can be calculated. The average value of the coordinate distance can be obtained to obtain the neighborhood radius, or the neighborhood radius can be obtained by weighted averaging. The business feature data that is farther away from the center of the circle has a smaller weight, and the business feature data that is closer to the center of the circle has a larger weight. The judgment criteria for farther and closer can be determined based on the type and quantity of specific business feature data.

[0083] According to an embodiment of the present disclosure, when merging sub-datasets, they can be sorted according to the size of the distance between the centers of the circles. First, the two sub-datasets with the smallest distance between the centers of the circles are merged to obtain a new sub-dataset. Then, all the sub-datasets are updated, and then the sub-dataset with the smallest distance between the center of the circle and the center of the circle of other sub-datasets is calculated and merged.

[0084] According to the embodiments of the present disclosure, compared with the traditional method of directly merging the original business feature data, the object of merging is a sub-dataset, which improves the processing efficiency of the business feature data. By merging the sub-datasets, the business feature data with consistent data structures or internal connections can be divided into the same target business data set, which helps to improve the accuracy of the suggestion generation model.

[0085] Figure 6 The following schematically shows a flow chart of a method for determining a target business data set according to an embodiment of the present disclosure. Figure 6 As shown, the target business data set determination method of this embodiment includes operations S610 to S650.

[0086] In operation S610 , the distance between the center of each sub-data set and the centers of the remaining sub-data sets is calculated.

[0087] In operation S620, for each pair of sub-data sets, when it is determined that the distance between the circle centers is less than the neighborhood radius, the two sub-data sets are merged to obtain a new sub-data set.

[0088] In operation S630, the center of the minimum circumscribed circle corresponding to the new sub-dataset is determined, and the distance between the center of the new sub-dataset and the centers of the remaining sub-datasets is calculated.

[0089] In operation S640, when it is determined that there are sub-datasets whose distance to the center of the new sub-dataset is smaller than the neighborhood radius among the remaining sub-datasets, the sub-datasets whose distance to the center is smaller than the neighborhood radius are iteratively merged based on the center of the new sub-dataset.

[0090] In operation S650, when it is determined that the distances between the center of the new sub-dataset and the centers of the remaining sub-datasets are greater than or equal to the neighborhood radius, the multiple sub-datasets obtained after merging are determined as the target business data set.

[0091] According to the embodiments of the present disclosure, the sub-datasets can be merged into the target business data set according to the idea of ​​DBSCAN. For example, the center of the minimum circumscribed circle corresponding to each sub-dataset is determined, so each sub-dataset is a sphere, which is a circle on a two-dimensional plane. Then, when merging sub-datasets, it is only necessary to calculate the distance between the centers of the circles. When the distance between the centers of the two sub-datasets is less than or equal to the neighborhood radius, the two sub-datasets are merged to obtain a new sub-dataset.

[0092] According to the embodiments of the present disclosure, the existing DBSCAN algorithm needs to search for the next core point for points other than the core point in each iteration, which will inevitably increase the computational complexity and reduce the computational efficiency. The reverse analysis method is used to divide the space from large clusters to small clusters to obtain sub-datasets, and then combined with the calculation ideas of DBSCAN, the clustering process is implemented to obtain the target business data set, avoiding the process of re-searching for the core point in each iteration. The object of the DBSCAN algorithm is the sub-dataset, which overcomes the problem of uncertain output results caused by the sensitivity of the DBSCAN algorithm to parameters in the prior art, improves the accuracy and efficiency of business feature data processing, and in addition, by merging the sub-datasets to obtain the target business data set, the laws and connections between the business feature data in the target business data set can be better obtained, and the target cluster model obtained by training the target business data set has higher accuracy.

[0093] According to an embodiment of the present disclosure, a data center point is determined by analyzing multiple business feature data, including: determining a data coordinate point of each business feature data in the data distribution space; and averaging multiple data coordinate points to obtain a data center point.

[0094] According to an embodiment of the present disclosure, the distribution of the business feature data is determined by determining the data coordinate points of each business feature data in the data distribution space, and no business feature data will be missed. When determining the data center point, it can be determined by calculating the mean of the data coordinate points or by determining the circumscribed circle of each business feature data in the data distribution space and then calculating the center of the circumscribed circle.

[0095] According to the embodiments of the present disclosure, in the process of determining the coordinates of the data center point, the position of each business feature data in the data distribution space is taken into consideration, and from a spatial perspective, the center point of all business feature data can be more accurately reflected.

[0096] Figure 7 The following schematically shows a process diagram of constructing a suggestion generation model according to an embodiment of the present disclosure. Figure 7 As shown, according to an embodiment of the present disclosure, the central server 302 uses the target business data set 703 to train multiple initial cluster models 702 respectively to obtain a suggestion generation model 701 composed of multiple trained cluster models 702, including: inputting multiple target business data sets 703 into the initial cluster models 702 respectively to train each initial cluster model 702 to obtain multiple trained cluster models 702. Determine the model parameters of each trained cluster model 702 to construct the suggestion generation model 701 according to the mean of multiple model parameters.

[0097] According to an embodiment of the present disclosure, an initial cluster model 702 is constructed for each target business data set 703, and each initial cluster model 702 is trained by applying the target business data set 703 to adjust the parameters of each initial cluster model 702, and finally a trained cluster model 702 is obtained, and the trained cluster models 702 are merged into a suggestion generation model 701, and the mean of the parameters of the trained cluster model 702 is used as the parameter of the suggestion generation model 701, so that the suggestion generation model 701 contains all data training results.

[0098] According to the embodiments of the present disclosure, the training method of the recommendation generation model first encrypts the business feature data when performing federated modeling. In the case of encryption, problems such as data inconsistency and data heterogeneity currently occur in the multi-party joint modeling process. Directly modeling the data will result in poor model effects and fail to meet business needs. However, by dividing the business feature data into different sub-data sets and building a model for each sub-data set, it is possible to build a model when the data features of each party are different, the distribution is different, and the data samples are inconsistent. By dividing the data features into different categories, data fusion and use are facilitated, and the problems of inaccurate federated learning models caused by not using clustering algorithms during federated modeling, which result in the independence of the data of each party and the inability to utilize the internal structures and laws of the data of different institutions, are overcome. Finally, the cluster model is aggregated into the recommendation generation model to achieve the construction of the recommendation generation model.

[0099] Figure 8 The flowchart of the model application method according to the embodiment of the present disclosure is schematically shown.

[0100] like Figure 8 As shown, the application method of this embodiment includes operations S810 and S820.

[0101] In operation S810, the service characteristic data sent by the participant is obtained.

[0102] In operation S820, the business feature data is input into the above-mentioned suggestion generation model for analysis, and the analysis result is output, wherein the analysis result is used to characterize the business processing suggestion corresponding to the business feature data, and the participating party can process the business according to the business processing suggestion.

[0103] The embodiments of the present disclosure, since the suggestion generation model is obtained by merging various cluster models into the suggestion generation model and taking the mean of the cluster model parameters as the suggestion generation model parameters, can meet the processing needs of business feature data of different participants, and the business feature data sent by the participants obtained in real time is input into the constructed suggestion generation model. The suggestion generation model can accurately analyze the business feature data sent by the participants and output more accurate analysis results, thereby improving the processing efficiency of the business feature data and the accuracy of the processing results.

[0104] Based on the above-mentioned training method of the suggestion generation model, the present disclosure also provides a training device for the suggestion generation model. Fig. 9 The device is described in detail.

[0105] Fig. 9 The structural block diagram of the training device of the suggestion generation model according to the embodiment of the present disclosure is schematically shown. Fig. 9 As shown, the training device 900 for the suggestion generation model of this embodiment includes: a data receiving module 910, a data clustering module 920, a model training module 930 and a model building module 940.

[0106] The data receiving module 910 is used to receive a plurality of service characteristic data sent by a plurality of participants to the central server for characterizing the service interaction between the plurality of participants. In one embodiment, the data receiving module 910 can be used to perform the operation S210 described above, which will not be described in detail here.

[0107] The data clustering module 920 is used to cluster the multiple business feature data used to characterize the business interaction between multiple participants using a clustering algorithm to obtain multiple target business data sets. In one embodiment, the data clustering module 920 can be used to perform the operation S220 described above, which will not be repeated here.

[0108] The model training module 930 is used to input multiple target business data sets into the corresponding initial cluster models respectively to train each initial cluster model to obtain multiple target cluster models. In one embodiment, the model training module 930 can be used to perform the operation S230 described above, which will not be repeated here.

[0109] The model building module 940 is used to train the untrained suggestion generation model using the federated learning algorithm according to the model parameters of each target cluster model to obtain a suggestion generation model for generating business processing suggestions for the participating party. In one embodiment, the model building module 940 can be used to perform the operation S240 described above, which will not be repeated here.

[0110] According to an embodiment of the present disclosure, the data clustering module 920 includes a data analysis submodule, a data partitioning submodule, and a data merging submodule.

[0111] The data analysis submodule is used to analyze multiple business feature data representing business interactions between multiple participants, wherein the data center point is the center coordinate point of the business feature data in the data distribution space.

[0112] The data partitioning submodule is used to partition multiple business feature data into multiple sub-data sets according to the distance between the data center point in the data distribution space and each business feature data, wherein the number of business feature data in each sub-data set is less than a preset number.

[0113] The data merging submodule is used to determine the center of the minimum circumscribed circle of each sub-data set in the data distribution space, so as to merge multiple sub-data sets according to the distance between the multiple centers to obtain multiple target business data sets, wherein the center distance between the multiple target business data sets is greater than the neighborhood radius.

[0114] According to an embodiment of the present disclosure, the data analysis submodule includes a coordinate determination unit and a center determination unit.

[0115] The coordinate determination unit is used to determine the data coordinate point of each service characteristic data in the data distribution space.

[0116] The center determination unit is used to calculate the average of multiple data coordinate points to obtain a data center point.

[0117] According to an embodiment of the present disclosure, the data partitioning submodule includes a data distribution unit, a distance determination unit, a calculation unit, a selection unit and a subset partitioning unit.

[0118] The data distribution unit is used to determine the data coordinate point farthest from the data center point as the first type of cluster point based on the data coordinate point of each business characteristic data in the data distribution space.

[0119] The distance determination unit is used to determine, from among the multiple data coordinate points, the data coordinate point farthest from the first type of cluster point as the second type of cluster point.

[0120] The calculation unit is used to calculate the vertical point from the center point of the data to the line connecting the first type of cluster points and the second type of cluster points.

[0121] The selection unit is used to determine the data coordinate point farthest from the vertical point from the multiple data coordinate points as the third type cluster point.

[0122] The subset division unit is used to calculate the distances between the remaining data coordinate points and the first type of cluster points, the second type of cluster points and the third type of cluster points, so as to divide the business feature data into multiple subset data sets according to the principle of the shortest distance.

[0123] According to an embodiment of the present disclosure, the subset partitioning unit includes an analysis subunit, a recursion subunit, a cluster point determination subunit and a partitioning subunit.

[0124] The analysis subunit is used to analyze multiple distances to divide the business feature data into sub-datasets corresponding to the first type of cluster points, sub-datasets corresponding to the second type of cluster points, and sub-datasets corresponding to the third type of cluster points according to the principle of closest distance.

[0125] The recursive subunit is used to repeatedly perform the following steps for each sub-data set, if it is determined that the amount of business feature data in the sub-data set is greater than or equal to the preset amount, to recursively divide the sub-data set until the amount of business feature data in the sub-data set is less than the preset amount.

[0126] The cluster point determination subunit is used to determine new first-class cluster points, new second-class cluster points and new third-class cluster points in the sub-dataset.

[0127] The sub-division unit is used to divide the business feature data in the sub-dataset according to the distances between the remaining data coordinate points in the sub-dataset and the new first-category cluster points, the new second-category cluster points and the new third-category cluster points, using the principle of the closest distance to obtain multiple new sub-datasets.

[0128] According to an embodiment of the present disclosure, the data merging submodule includes an interval determining unit, a mean value calculating unit and a merging unit.

[0129] The interval determination unit is used to determine the coordinate distance between each service feature data and the remaining service feature data in the data distribution space.

[0130] The mean calculation unit is used to calculate the average value of multiple coordinate distances to obtain the neighborhood radius.

[0131] The merging unit is used to merge sub-data sets whose center distances are smaller than the neighborhood radius in pairs to obtain multiple target business data sets.

[0132] According to an embodiment of the present disclosure, the merging unit includes a distance calculation subunit, a data set aggregation subunit, a distance calculation subunit, a data set iteration subunit and a target determination subunit.

[0133] The distance calculation subunit is used to calculate the distance between the center of each sub-dataset and the centers of the other sub-datasets.

[0134] The data set aggregation subunit is used to merge the two sub data sets for each pair of sub data sets when it is determined that the distance between the circle centers is less than the neighborhood radius to obtain a new sub data set.

[0135] The distance calculation subunit is used to determine the center of the minimum circumscribed circle corresponding to the new sub-dataset and calculate the distance between the center of the new sub-dataset and the centers of the remaining sub-datasets.

[0136] The data set iteration subunit is used to iteratively merge the sub data sets whose center distances are less than the neighborhood radius based on the center of the new sub data set when it is determined that there are sub data sets in the remaining sub data sets whose center distances are less than the neighborhood radius.

[0137] The target determination subunit is used to determine the multiple sub-datasets obtained after merging as the target business data set when it is determined that the distances between the center of the new sub-dataset and the centers of the other sub-datasets are greater than or equal to the neighborhood radius.

[0138] Fig.10 The structure block diagram of the model application device according to the embodiment of the present disclosure is schematically shown. Fig.10 As shown, the model application device 1000 of this embodiment includes: a data acquisition module 1010 and a result output module 1020.

[0139] The data acquisition module 1010 is used to acquire the business characteristic data sent by the participants.

[0140] The result output module 1020 is used to input the business feature data into the above-mentioned suggestion generation model for analysis and output the analysis results, wherein the analysis results are used to characterize the business processing suggestions corresponding to the business feature data, and the participants can process the business according to the business processing suggestions.

[0141] According to an embodiment of the present disclosure, any multiple modules in the data receiving module 910, the data clustering module 9 / 20, the model training module 930 and the model building module 940, and the data acquisition module 1010 and the result output module 1020 can be combined in one module for implementation, or any one of the modules can be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules can be combined with at least part of the functions of other modules and implemented in one module. According to an embodiment of the present disclosure, at least one of the data receiving module 910, the data clustering module 9 / 20, the model training module 930 and the model building module 940, and the data acquisition module 1010 and the result output module 1020 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application specific integrated circuit (ASIC), or can be implemented by hardware or firmware such as any other reasonable way of integrating or packaging the circuit, or implemented in any one of the three implementation methods of software, hardware and firmware or in an appropriate combination of any of them. Alternatively, at least one of the data receiving module 910, the data clustering module 9 / 20, the model training module 930 and the model building module 940 and the data acquisition module 1010 and the result output module 1020 can be at least partially implemented as a computer program module, which can perform corresponding functions when the computer program module is executed.

[0142] Fig.11 A block diagram of an electronic device suitable for implementing a training method for a suggestion generation model and a method for applying the model according to an embodiment of the present disclosure is schematically shown.

[0143] like Fig.11 As shown, the electronic device 1100 according to an embodiment of the present disclosure includes a processor 1101, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1102 or a program loaded from a storage part 1108 to a random access memory (RAM) 1103. The processor 1101 may include, for example, a general-purpose microprocessor (such as a CPU), an instruction set processor and / or a related chipset and / or a special-purpose microprocessor (for example, an application-specific integrated circuit (ASIC)), etc. The processor 1101 may also include an onboard memory for caching purposes. The processor 1101 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present disclosure.

[0144] In RAM 1103, various programs and data required for the operation of electronic device 1100 are stored. Processor 1101, ROM 1102 and RAM 1103 are connected to each other through bus 1104. Processor 1101 performs various operations of the method flow according to the embodiment of the present disclosure by executing the program in ROM 1102 and / or RAM 1103. It should be noted that the program can also be stored in one or more memories other than ROM 1102 and RAM 1103. Processor 1101 can also perform various operations of the method flow according to the embodiment of the present disclosure by executing the program stored in the one or more memories.

[0145] According to an embodiment of the present disclosure, the electronic device 1100 may further include an input / output (I / O) interface 1105, which is also connected to the bus 1104. The electronic device 1100 may further include one or more of the following components connected to the input / output (I / O) interface 1105: an input portion 1106 including a keyboard, a mouse, etc.; an output portion 1107 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage portion 1108 including a hard disk, etc.; and a communication portion 1109 including a network interface card such as a LAN card, a modem, etc. The communication portion 1109 performs communication processing via a network such as the Internet. A drive 1110 is also connected to the input / output (I / O) interface 1105 as needed. A removable medium 1111, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 1110 as needed, so that a computer program read therefrom is installed into the storage portion 1108 as needed.

[0146] The present disclosure also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or may exist independently without being assembled into the device / apparatus / system. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the method according to the embodiment of the present disclosure is implemented.

[0147] According to an embodiment of the present disclosure, a computer-readable storage medium may be a non-volatile computer-readable storage medium, for example, may include but is not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, an apparatus, or a device. For example, according to an embodiment of the present disclosure, a computer-readable storage medium may include the ROM 1102 and / or RAM 1103 described above and / or one or more memories other than ROM 1102 and RAM 1103.

[0148] The embodiment of the present disclosure also includes a computer program product, which includes a computer program, and the computer program includes a program code for executing the method shown in the flowchart. When the computer program product is run in a computer system, the program code is used to enable the computer system to implement the training method for implementing the suggestion generation model and the application method of the model provided by the embodiment of the present disclosure.

[0149] The above functions defined in the system / device of the embodiment of the present disclosure are performed when the computer program is executed by the processor 1101. According to the embodiment of the present disclosure, the system, device, module, unit, etc. described above can be implemented by a computer program module.

[0150] In one embodiment, the computer program may rely on tangible storage media such as optical storage devices, magnetic storage devices, etc. In another embodiment, the computer program may also be transmitted and distributed in the form of signals on a network medium, and downloaded and installed through the communication part 1109, and / or installed from a removable medium 1111. The program code contained in the computer program may be transmitted using any appropriate network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0151] In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 1109, and / or installed from the removable medium 1111. When the computer program is executed by the processor 1101, the above functions defined in the system of the embodiment of the present disclosure are performed. According to the embodiment of the present disclosure, the system, device, means, module, unit, etc. described above can be implemented by a computer program module.

[0152] According to an embodiment of the present disclosure, the program code for executing the computer program provided by the embodiment of the present disclosure can be written in any combination of one or more programming languages. Specifically, these computing programs can be implemented using high-level process and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, Java, C++, python, "C" language or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, partially on the remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (for example, using an Internet service provider to connect through the Internet).

[0153] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0154] It will be appreciated by those skilled in the art that the features described in the various embodiments of the present disclosure may be combined and / or combined in a variety of ways, even if such combinations or combinations are not explicitly described in the present disclosure. In particular, without departing from the spirit and teachings of the present disclosure, the features described in the various embodiments of the present disclosure may be combined and / or combined in a variety of ways. All of these combinations and / or combinations fall within the scope of the present disclosure.

[0155] The embodiments of the present disclosure are described above. However, these embodiments are only for illustrative purposes and are not intended to limit the scope of the present disclosure. Although the embodiments are described above, this does not mean that the measures in the various embodiments cannot be used in combination to advantage. Without departing from the scope of the present disclosure, those skilled in the art may make a variety of substitutions and modifications, which should all fall within the scope of the present disclosure.

Claims

1. A training method for a suggestion generation model, characterized in that: The method comprises: Receiving a plurality of business characteristic data sent by a plurality of participants to a central server for characterizing business interactions between the plurality of participants; Using a clustering algorithm, clustering is performed on the plurality of business feature data used to characterize business interactions between the plurality of participants to obtain a plurality of target business data sets; Inputting the plurality of target business data sets into the corresponding initial cluster models respectively, so as to train each of the initial cluster models and obtain a plurality of target cluster models; According to the model parameters of each of the target cluster models, a federated learning algorithm is used to train an untrained suggestion generation model to obtain a suggestion generation model for generating business processing suggestions for the participating party.

2. The method according to claim 1, characterized in that The clustering algorithm is used to cluster the multiple business feature data used to characterize the business interaction between the multiple participants to obtain multiple target business data sets, including: By analyzing the plurality of service feature data, a data center point is determined, wherein the data center point is a center coordinate point of the plurality of service feature data in a data distribution space; Dividing the plurality of service feature data into a plurality of sub-data sets according to the distance between the central point of the data in the data distribution space and each of the service feature data, wherein the number of the service feature data in each of the sub-data sets is less than a preset number; Determine the center of the minimum circumscribed circle of each sub-data set in the data distribution space, and merge multiple sub-data sets according to the distance between the multiple centers to obtain multiple target business data sets, wherein the distance between the centers of the multiple target business data sets is greater than the neighborhood radius.

3. The method according to claim 2, characterized in that The dividing the plurality of service feature data into a plurality of sub-data sets according to the distance between the central point of the data in the data distribution space and each of the service feature data comprises: Based on the data coordinate points of each of the service feature data in the data distribution space, the data coordinate point farthest from the data center point is determined as a first type of cluster point; From the plurality of data coordinate points, the data coordinate point farthest from the first type of cluster point is determined as the second type of cluster point: Calculate the vertical point from the data center point to the line connecting the first type of cluster points and the second type of cluster points; From the plurality of data coordinate points, the data coordinate point farthest from the vertical point is determined as a third type cluster point: The distances between the remaining data coordinate points and the first type of cluster points, the second type of cluster points and the third type of cluster points are calculated respectively, so as to divide the service feature data into a plurality of sub-data sets according to the principle of the shortest distance.

4. The method according to claim 3, characterized in that The dividing the service characteristic data into a plurality of sub-data sets according to the principle of closest distance includes: Analyzing the plurality of distances, so as to divide the service characteristic data into sub-data sets corresponding to the first type of cluster points, sub-data sets corresponding to the second type of cluster points, and sub-data sets corresponding to the third type of cluster points according to a principle of closest distance; For each of the sub-data sets, when it is determined that the amount of the service feature data in the sub-data set is greater than or equal to the preset amount, repeatedly performing the following steps to recursively divide the sub-data set until the amount of the service feature data in the sub-data set is less than the preset amount: Determining new first-category cluster points, new second-category cluster points, and new third-category cluster points in the sub-dataset; According to the distances between the remaining data coordinate points in the sub-dataset and the new first-category cluster points, the new second-category cluster points and the new third-category cluster points, the business feature data in the sub-dataset is divided using the closest distance principle to obtain multiple new sub-datasets.

5. The method according to claim 2, characterized in that: The merging of the plurality of sub-data sets according to the distances between the plurality of circle centers to obtain a plurality of target business data sets includes: In the data distribution space, determining the coordinate distance between each of the service feature data and the remaining service feature data; averaging a plurality of the coordinate distances to obtain the neighborhood radius; The sub-data sets whose center distances are smaller than the neighborhood radius are merged in pairs to obtain a plurality of the target business data sets.

6. The method according to claim 5, characterized in that The step of merging two by two of the sub-data sets whose centers are less than the neighborhood radius to obtain a plurality of the target business data sets includes: Calculating the distance between the center of each sub-dataset and the centers of the remaining sub-datasets; For each pair of the sub-datasets, when it is determined that the distance between the circle centers is less than the neighborhood radius, the two sub-datasets are merged to obtain a new sub-dataset; Determine the center of the minimum circumscribed circle corresponding to the new sub-dataset, and calculate the distance between the center of the new sub-dataset and the centers of the remaining sub-datasets; When it is determined that there is a sub-dataset in the remaining sub-datasets whose distance to the center of the new sub-dataset is smaller than the neighborhood radius, iteratively merging the sub-datasets whose distance to the center of the new sub-dataset is smaller than the neighborhood radius based on the center of the new sub-dataset; When it is determined that the distances between the center of the new sub-dataset and the centers of the other sub-datasets are greater than or equal to the neighborhood radius, the multiple sub-datasets obtained after merging are determined as the target business data set.

7. The method according to claim 2, characterized in that: The step of analyzing the plurality of service characteristic data to determine the data center point includes: Determine a data coordinate point of each of the service characteristic data in the data distribution space; The average of the multiple data coordinate points is calculated to obtain the data center point.

8. A method for applying a model, characterized in that: The method comprises: Obtain business characteristic data sent by the participants; The business characteristic data is input into the suggestion generation model described in any one of claims 1 to 7 for analysis, and the analysis result is output, wherein the analysis result is used to characterize the business processing suggestion corresponding to the business characteristic data, and the participant can process the business according to the business processing suggestion.

9. A training device for a suggestion generation model, characterized in that: The device comprises: A data receiving module is used to receive multiple business feature data sent by multiple participants to the central server to characterize the business interaction between the multiple participants. A data clustering module, used for clustering the plurality of business feature data used to characterize the business interaction between the plurality of participants using a clustering algorithm to obtain a plurality of target business data sets; A model training module, used for inputting the plurality of target business data sets into the corresponding initial cluster models respectively, so as to train each of the initial cluster models and obtain a plurality of target cluster models; The model building module is used to train an untrained suggestion generation model using a federated learning algorithm according to the model parameters of each target cluster model to obtain a suggestion generation model for generating business processing suggestions for the participating party.

10. An electronic device, comprising: one or more processors; a memory for storing one or more computer programs, It is characterized in that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 7 or claim 8.

11. A computer-readable storage medium having a computer program or instruction stored thereon, characterized in that: When the computer program or instruction is executed by a processor, the steps of the method according to any one of claims 1 to 7 or claim 8 are implemented.

12. A computer program product comprising a computer program or instructions, characterized in that When the computer program or instruction is executed by a processor, the steps of the method according to any one of claims 1 to 7 or claim 8 are implemented.

Images