Federated Modeling Method, Device, Medium and Program Product for Feature Binning Model

Abstract

The present application discloses a federated modeling method, device, medium and program product for a feature binning model, which relates to the field of artificial intelligence. The method includes: receiving different model information sent by different federated participants after respectively establishing an interactive feature binning model; merging the different model information to obtain an aggregated feature binning model, and distributing the aggregated feature binning model to different federated participants, so that the corresponding different federated participants receive the aggregated feature binning model and feedback whether a preset federated training end condition is reached based on the aggregated feature binning model; if feedback information indicating that the preset federated training end condition is not reached is received, returning to the step of receiving different model information sent by different federated participants after respectively establishing an interactive feature binning model, until feedback information indicating that the preset federated training end condition is reached is received from the corresponding federated participant, so as to obtain a target federated feature binning model. The present invention can improve the modeling efficiency and ensure data privacy and security.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence, and in particular, to a federated modeling method, device, computer-readable storage medium, and computer program product for a feature binning model. Background Art

[0002] With the continuous development of machine learning technology, more and more fields have begun to apply machine learning technology. In machine learning technology, the robustness of machine learning models is crucial, and binning processing for interval features can effectively enhance the robustness of machine learning models. Binning processing is a feature processing method that divides interval feature values into multiple discrete values, which can make the feature values more stable, thereby improving the robustness of machine learning models.

[0003] Currently, common feature binning processing includes equal-frequency, equal-distance, and optimal binning for the obtained modeling data. However, after these feature binning processes default to automatically running, manual interactive adjustment by humans is usually required. Thus, it is difficult to avoid affecting the privacy and security of data. Moreover, with the increasing attention to data privacy and security in recent years and the release of corresponding policies and regulations, the traditional method of obtaining modeling data in one place, binning the data, and then establishing a machine learning model is no longer convenient and feasible.

[0004] In summary, how to establish a feature binning model for binning modeling data while ensuring data privacy and security, improving the model construction efficiency, as well as the accuracy and robustness of the model, has clearly become a technical problem that urgently needs to be solved in the industry. Summary of the Invention

[0005] The main purpose of this application is to provide a federated modeling method, device, storage medium, and program product for a feature binning model, aiming to establish a feature binning model for binning modeling data based on multi-person cooperation, so as to improve the modeling efficiency, as well as the accuracy and robustness of the feature binning model while ensuring data privacy and security.

[0006] To achieve the above objective, this application provides a federated modeling method for a feature binning model, which is applied to a server. The federated interactive modeling method of the decision tree model includes:

[0007] Receiving different model information sent by different federated participants after they respectively establish interactive feature binning models;

[0008] Merging the different model information to obtain an aggregated feature binning model, and distributing the aggregated feature binning model to the different federated participants, so that the corresponding different federated participants receive the aggregated feature binning model and feedback whether the preset federated training end condition is reached based on the aggregated feature binning model;

[0009] If feedback information indicating that a preset federal training end condition is not met is received, return to the step of receiving different model information sent after different federated participants respectively establish interactive feature binning models until feedback information indicating that the preset federal training end condition is met is received from the corresponding federated participant, so as to obtain a target federated feature binning model.

[0010] Optionally, the step of combining the different model information to obtain an aggregated feature binning model includes:

[0011] Based on the different model information, determine a first model that performs binning processing on data features and a second model that does not perform binning processing on the data features in the interactive feature binning models of different federated participants;

[0012] Wherein, the data features are the data features of the modeling data used when different federated participants establish the interactive feature binning models;

[0013] Determine the first model as the aggregated feature binning model.

[0014] Optionally, the step of combining the different model information to obtain an aggregated feature binning model further includes:

[0015] Based on the different model information, determine whether the feature binning rules run when different federated participants establish interactive feature binning models are the same;

[0016] If the feature binning rules are different, combine the feature binning rules to obtain an aggregated feature binning model.

[0017] Optionally, the step of distributing the aggregated feature binning model to the different federated participants for the corresponding different federated participants to receive the aggregated feature binning model and feedback whether the preset federal training end condition is met includes:

[0018] Distribute the aggregated feature binning model to the different federated participants so that after the corresponding different federated participants receive the aggregated feature binning model, calculate the model metrics of the aggregated feature binning model and feedback whether the preset federal training end condition is met, where the preset federal training end condition includes reaching a preset model metric threshold or reaching a preset maximum number of federal training iteration rounds.

[0019] Optionally, the step of receiving different model information sent after different federated participants respectively establish interactive feature binning models includes:

[0020] Receive different model information sent by different federal participants after establishing interactive feature binning models respectively. Among them, the different federal participants obtain a feature binning model to be operated and an operation task for operating the feature binning model to be operated, and operate the feature binning model to be operated according to the operation task to obtain an interactive feature binning model, and determine model information according to the interactive feature binning model.

[0021] Optionally, after the step of returning to receive different model information sent by different federal participants after establishing interactive feature binning models respectively until the feedback information indicating that the preset federal training end condition is reached is received from the corresponding federal participant to obtain the target federal feature binning model, the method includes:

[0022] Obtain data to be processed, and input the data to be processed into the target federal feature binning model;

[0023] Perform binning processing on the data to be processed based on the target federal feature binning model to obtain binned data, and establish a machine learning model based on the binned data.

[0024] In addition, to achieve the above object, the present application also provides a federal modeling method for a feature binning model, which is applied to different federal participants in the federal modeling method for a feature binning model as described above. The federal modeling method for a feature binning model includes:

[0025] After establishing an interactive feature binning model, upload the model information of the interactive feature binning model to the server;

[0026] Receive the aggregated feature binning model obtained by the server by merging the model information, and feedback whether the preset federal training end condition is reached based on the aggregated feature binning model;

[0027] If the feedback information indicating that the preset federal training end condition is not reached is sent to the server, return to the step of uploading the model information of the interactive feature binning model to the server after establishing the interactive feature binning model until the feedback information indicating that the preset federal training end condition is reached is sent to the server for the server to obtain the target federal feature binning model.

[0028] In addition, to achieve the above object, the present application also provides a federal modeling device for a feature binning model, which is applied to a server. The federal interactive modeling device for a random forest model includes:

[0029] A first receiving module, configured to receive different model information sent by different federal participants after establishing interactive feature binning models respectively;

[0030] A merging and distributing module, configured to merge the different model information to obtain an aggregated feature binning model, and distribute the aggregated feature binning model to the different federal participants, so that the corresponding different federal participants receive the aggregated feature binning model and feedback whether a preset federal training end condition is reached based on the aggregated feature binning model;

[0031] A first judgment module, configured to, if feedback information indicating that the preset federal training end condition is not reached is received, return to the step of receiving different model information sent after different federal participants respectively establish interactive feature binning models, until feedback information indicating that the preset federal training end condition is reached is fed back by the corresponding federal participant, so as to obtain a target federal feature binning model;

[0032] Wherein, the federal interactive modeling device of the random forest model is also applied to different federal participants connected to the server, and the federal interactive modeling device of the random forest model further includes:

[0033] A model construction module, configured to upload model information of the interactive feature binning model to the server after establishing the interactive feature binning model;

[0034] A second receiving module, configured to receive the aggregated feature binning model obtained by the server merging the model information, and feedback whether a preset federal training end condition is reached based on the aggregated feature binning model;

[0035] A second judgment module, configured to, if feedback information indicating that the preset federal training end condition is not reached is fed back to the server, return to the step of uploading the model information of the interactive feature binning model to the server after establishing the interactive feature binning model, until feedback information indicating that the preset federal training end condition is reached is fed back to the server, so that the server can obtain a target federal feature binning model.

[0036] The present application further provides a federal modeling device for a feature binning model. The federal modeling device for the feature binning model is an entity device, and the federal modeling device for the feature binning model includes: a memory, a processor, and a program of the federal modeling method for the feature binning model stored on the memory and executable on the processor. When the program of the federal modeling method for the feature binning model is executed by the processor, the steps of the federal modeling method for the feature binning model as described above can be implemented.

[0037] The present application further provides a computer-readable storage medium, on which a program for implementing the federal modeling method for the feature binning model is stored. When the program for the federal modeling method for the feature binning model is executed by a processor, the steps of the federal modeling method for the feature binning model as described above are implemented.

[0038] The present application also provides a computer program product, including a computer program, which when executed by a processor implements the steps of the method for federated modeling of the feature binning model described in any one of the above.

[0039] A method, device, equipment, computer-readable storage medium and computer program product for federated modeling of a feature binning model provided by the present application, wherein the server receives different model information sent by different federated participants after respectively establishing an interactive feature binning model; combines the different model information to obtain an aggregated feature binning model, and distributes the aggregated feature binning model to the different federated participants, so that the corresponding different federated participants receive the aggregated feature binning model and feedback whether a preset federated training end condition is reached based on the aggregated feature binning model; if feedback information indicating that the preset federated training end condition is not reached is received, returns to the step of receiving the different model information sent by different federated participants after respectively establishing an interactive feature binning model, until feedback information indicating that the preset federated training end condition is reached is received from the corresponding federated participant, so as to obtain a target federated feature binning model.

[0040] The present application realizes the joint modeling of an interactive feature binning model by multiple federated participants, enabling the modeling process of the interactive feature binning model to use more training data on the basis of ensuring data privacy and security, thereby improving the construction efficiency of the feature binning model, ensuring the accuracy and robustness of the model, and ensuring the privacy and security of the modeling data. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The drawings herein are incorporated into the specification and form a part of the specification, showing embodiments consistent with the present application and used together with the specification to explain the principles of the present application.

[0042] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, for those of ordinary skill in the art, other drawings can also be obtained based on these drawings without creative efforts.

[0043] Figure 1 It is a schematic flowchart of an embodiment of the method for federated modeling of the feature binning model of the present application;

[0044] Figure 2 It is a schematic flowchart of another embodiment of the method for federated modeling of the feature binning model of the present application;

[0045] Figure 3 It is a schematic diagram of the device structure of the hardware operating environment involved in the solution of the embodiment of the present application.

[0046] The realization of the purpose, functional features and advantages of this application will be further described in conjunction with the embodiments with reference to the accompanying drawings. Specific Embodiments

[0047] It should be understood that the specific embodiments described herein are only used to explain this application and are not used to limit this application.

[0048] The embodiments of this application provide a federated modeling method for a feature binning model. In the first embodiment of the federated modeling method for the feature binning model of this application, refer to Figure 1 , the federated modeling method for the feature binning model includes:

[0049] Step S10, receive different model information sent by different federated participants after respectively establishing an interactive feature binning model;

[0050] Step S20, merge the different model information to obtain an aggregated feature binning model, and distribute the aggregated feature binning model to the different federated participants, so that the corresponding different federated participants receive the aggregated feature binning model, and based on the aggregated feature binning model, feedback whether the preset federated training end condition is reached;

[0051] Step S30, if feedback information indicating that the preset federated training end condition is not reached is received, return to the step of receiving different model information sent by different federated participants after respectively establishing an interactive feature binning model, until the feedback information indicating that the preset federated training end condition is reached is fed back by the corresponding federated participant, so as to obtain a target federated feature binning model.

[0052] The specific steps are as follows:

[0053] Step S10, receive different model information sent by different federated participants after respectively establishing an interactive feature binning model;

[0054] It should be noted that in this embodiment, the federated modeling method for the feature binning model is applied to a federated modeling device for the feature binning model, and the federated modeling device for the feature binning model belongs to a federated modeling device for the feature binning model. Currently, with the continuous development of machine learning technology, more and more fields begin to apply machine learning technology. In machine learning technology, the robustness of machine learning models is crucial, and the binning process for interval features can effectively enhance the robustness of machine learning models. The binning process is a feature processing method that divides interval feature values into multiple discrete values, which can make the feature values more stable, thereby improving the robustness of machine learning models.

[0055] Currently, common feature binning processes include equal frequency, equal distance, and optimal binning for the obtained modeling data. However, after these feature binning processes are automatically run by default, manual interactive adjustments usually need to be made by humans. Thus, it is difficult to avoid affecting data privacy and security. Moreover, with the increasing attention to data privacy and security in recent years and the release of corresponding policies and regulations, the traditional approach of obtaining modeling data in one place, binning the data, and then building a machine learning model is no longer convenient and feasible.

[0056] In summary, how to build a feature binning model for binning modeling data while ensuring data privacy and security, improving model construction efficiency, as well as the accuracy and robustness of the model, has become an urgent technical problem in the industry.

[0057] In addition, it should be noted that in this embodiment, the interactive feature binning model can include three levels of interaction methods or three levels of interaction meanings:

[0058] Interaction method 1: It means that the user can operate on the feature binning model displayed in the visual interactive interface to trigger an interactive operation instruction. Then, based on the interactive operation instruction, an interactive feature binning model is obtained. Among them, the feature binning model displayed in the visual interactive interface can be one or more;

[0059] Interaction method 2: Through manual intervention or human interaction, conventional interactive operations such as retaining, deleting, and screening can be performed on the data features for binning the displayed feature binning model;

[0060] Interaction method 3: Through manual intervention or human interaction, operations such as adjusting the feature binning rules can be performed on the displayed feature binning model.

[0061] It should be noted that in this embodiment, the application scenario of the federated modeling method of the feature binning model can be: a feature binning model that needs to be effectively predicted and is subject to manual intervention. In order to improve the generalization ability of the model, during the modeling process, the personnel participating in the modeling are divided into multiple groups, that is, through multiple groups of manual interventions. Each participating party obtains an interactive feature binning model corresponding to its group based on the intervention method of its group and the data within its group. Then, the feature binning models obtained by federating or combining multiple groups are used to obtain the target federated feature binning model. It should be emphasized that during the local iterative training process of the participating parties, manual intervention is carried out, such as the selection of manual intervention reference factors or the selection of the reference factor range.

[0062] It should be noted that in this embodiment, the application scenario of the federated modeling method for the feature binning model can also be: it is necessary to establish an effective prediction and a classification feature binning model with manual intervention. In order to improve the generalization ability of the model, therefore, during the modeling process, the people participating in the modeling are divided into multiple groups, that is, through multiple groups of manual interventions. Each participating party obtains the classification feature binning model corresponding to its group based on the intervention method of its group and the classification data within its group. Furthermore, the classification feature binning models obtained from multiple groups are federated or combined to obtain the target classification federated feature binning model.

[0063] Specifically, each participating party has a classification model dataset: X, where X contains n data items x1, x2, x3,..., xi,..., xn. Each data item xi contains k values (xi1, xi2, xi3, xi4,..., xik). The main participating party's corresponding model dataset has a classification label Y: Y contains n data items y1, y2, y3, y4,..., yi, etc. For the classification model (binary classification or multi-classification), yi is discrete.

[0064] Alternatively, the application scenario of the federated modeling method for the feature binning model of this application can also be: it is necessary to establish an effective prediction and a regression feature binning model with manual intervention. In order to improve the generalization ability of the model, therefore, during the modeling process, the people participating in the modeling are divided into multiple groups, that is, through multiple groups of manual interventions. Each participating party obtains the regression feature binning model corresponding to its group based on the intervention method of its group and the regression data within its group. Furthermore, the regression feature binning models obtained from multiple groups are federated or combined to obtain the target federated regression feature binning model.

[0065] Specifically, each participating party has a model dataset: X, where X contains n data items x1, x2, x3,..., xi,..., xn. Each data item xi contains k values (xi1, xi2, xi3, xi4,..., xik). In this embodiment, the model dataset has a corresponding label Y: Y contains n data items y1, y2, y3, y4,..., yi, etc. For the regression model (binary classification or multi-classification), yi is continuous.

[0066] Receive different model information sent by different federal participants after they respectively establish interactive random forest models. That is, after different federal participants respectively establish interactive feature binning models through interactive operations, they respectively send the corresponding local model information to the server. Among them, the model information can be information such as model parameters. Specifically, the model parameters can be the feature binning rules run locally when establishing the interactive feature binning model, the feature types of the data features for which feature binning is performed, and the number of features, etc. Specifically, after Party A establishes interactive feature binning model 1, it sends model information 1 corresponding to interactive random feature binning 1 to the server. After Party B establishes interactive feature binning model 2, it also sends model information 2 corresponding to interactive feature binning model 2 to the server.

[0067] Further, in a feasible embodiment, step S10, receiving different model information sent by different federal participants after they respectively establish interactive feature binning models, may include:

[0068] Step S11, receive different model information sent by different federal participants after they respectively establish interactive feature binning models;

[0069] It should be noted that in this embodiment, the different federal participants obtain the feature binning model to be operated and the operation tasks for operating the feature binning model to be operated, and operate the feature binning model to be operated according to the operation tasks to obtain an interactive feature binning model, and determine model information according to the interactive feature binning model.

[0070] In addition, in this embodiment, the manner in which each participant establishes an interactive feature binning model is described. Specifically, the training data for training the interactive feature binning model is respectively stored in multiple participants. That is, each participant locally stores some training data. The server side can be pre-connected to multiple participants, or the server side can establish a connection with the participant when the participant requests a connection. It should be noted that this embodiment does not limit the organization method of each participant.

[0071] During the interactive modeling process of each participant, the user of the corresponding participant can trigger an interactive operation instruction by operating the feature binning model displayed in the visual output interactive operation interface. The participant can determine the feature binning model to be operated and the operation task for operating the feature binning model based on the interactive operation instruction. Among them, according to different interactive operation instructions triggered by the user, the operation tasks are different. For example, if the user wants to delete a certain feature binning rule for the feature binning model displayed on the current interface and triggers an interactive operation instruction through the interactive interface, the feature binning model to be operated obtained by the participant is the feature binning model displayed on the current interface, and the operation task is to delete a certain feature binning rule for the feature binning model. In order to establish a unified federated model, each participant user needs to determine the corresponding and consistent feature binning model to be operated and the operation task.

[0072] Of course, during the interactive modeling process of each participant, in order to ensure the unity of federated modeling, the user corresponding to the server can trigger an interactive operation instruction by operating the feature binning model displayed in the visual output interactive interface, and then determine the feature binning model to be operated and the operation task for operating the feature binning model based on the interactive operation instruction. Furthermore, the server distributes the obtained feature binning model and the corresponding operation task to each participant. Among them, according to different interactive operation instructions triggered by the user, the operation tasks are also different.

[0073] Specifically, in this embodiment, the operation tasks may include conventional interactive operations such as retaining, deleting, and / or adding the data features for feature binning processing of the feature binning model, and may also include interactive operations such as feature binning rule adjustment. According to different operation tasks, the participant performs different operations on the feature binning model. For example, when the operation task is to perform feature binning rule adjustment, the participant needs to determine the feature binning rule run when establishing the feature binning model, and then make corresponding adjustments to the feature binning rule according to the adjustment content.

[0074] After each participant performs local operations, i.e., local training, on the feature binning model, the participant determines the model information of the federation based on the feature binning model after the operation. Specifically, each participant continuously performs local iterative training on the feature binning model after the operation with local training data, and during the iterative training process, obtains the model information after manual intervention. For example, the model information after manual intervention includes the number of data features and the names of data features for feature binning processing by the feature binning model. When the local training is completed, the model information of the trained interactive feature binning model is obtained.

[0075] Specifically, for example, after Party A finishes training, the model information of the interactive feature binning model 1 includes: a model trained based on running the feature binning rule 1, and this model performs feature binning processing on 3 data features, and the names of these 3 data features are age, income, and education level respectively. Or, for example, after Party B finishes training, the model information of the interactive feature binning model 2 includes: a model trained based on running the feature binning rule 2, and this model performs feature binning processing on 5 data features, and the names of these 3 data features are age, household register, income, position, and credit respectively.

[0076] Step S20: Merge the different model information to obtain an aggregated feature binning model, and distribute the aggregated feature binning model to the different federal participants, so that the corresponding different federal participants receive the aggregated feature binning model and feedback whether the preset federal training end condition is met based on the aggregated feature binning model;

[0077] In this embodiment, after the server obtains the different model information of the different federal participants, it merges the different model information to obtain an aggregated feature binning model.

[0078] It should be noted that in this embodiment, the ways for the server to merge the different model information to obtain an aggregated feature binning model include:

[0079] Method 1: The method of selecting a model.

[0080] Further, in a feasible embodiment, in the above step S20, the step of "merging the different model information to obtain an aggregated feature binning model" may include:

[0081] In a feasible embodiment, in the above step S20, the step of "merging the different model information to obtain an aggregated random forest model" may include:

[0082] Step S21: Based on the different model information, determine a first model that performs binning processing on data features and a second model that does not perform binning processing on the data features in the interactive feature binning models of different federal participants;

[0083] Step S22: Determine the first model as the aggregated feature binning model.

[0084] It should be noted that in this embodiment, the data features are the data features of the modeling data used by the different federal participants to establish the interactive feature binning model;

[0085] The server determines that the interactive feature binning models established locally by different federal participants are data features of the modeling data used, and then detects the first model that performs feature binning processing on any one of the data features when the different federal participants establish the interactive feature binning models respectively, and, detects the second model that does not perform feature binning processing on any one of the data features when the different federal participants establish the interactive feature binning models respectively. Thus, the server directly determines the first model as the aggregated feature binning model.

[0086] Specifically, for example, after the server receives the model information of the interactive feature binning model 1 after A participant's training is completed, and the model information of the interactive feature binning model 2 after B participant's training is completed, the server further based on the model information of the interactive feature binning model 1: The model is trained based on running the feature binning rule 1, and this model performs feature binning processing on 3 data features, and the names of these 3 data features are age, income, and education level respectively. It is detected that when A participant establishes the interactive feature binning model 1 locally, it does not perform feature binning processing on the data feature - age. Thus, the server determines that the interactive feature binning model 1 is the second model. In addition, the server further based on the model information of the interactive feature binning model 2: The model is trained based on running the feature binning rule 2, and this model performs feature binning processing on 5 data features, and the names of these 3 data features are age, household register, income, position, and credit respectively. It is detected that when B participant establishes the interactive feature binning model 2 locally, it performs feature binning processing on all 5 data features. Thus, the server determines that the interactive feature binning model 2 is the first model. In this way, the server directly uses the first model - the interactive feature binning model 2 as the aggregated feature binning model.

[0087] It should be noted that in this embodiment, the way for the server to merge different model information to obtain the aggregated feature binning model also includes:

[0088] Way two: The way of taking the union.

[0089] In a feasible embodiment, in the above step S20, the step of "merging the different model information to obtain the aggregated random forest model" may further include:

[0090] Step S23, determining whether the feature binning rules run when different federal participants establish the interactive feature binning models are the same based on the different model information;

[0091] Step S24, if the feature binning rules are different, then merge the feature binning rules to obtain the aggregated feature binning model.

[0092] Specifically, for example, the server receives the model information of the interactive feature binning model 1 after Party A's training is completed: the model is trained based on running the feature binning rule 1, and this model performs feature binning processing on 3 data features, and the names of these 3 data features are age, income, and education level respectively. And the server receives the model information of the interactive feature binning model 2 after Party B's training is completed: the model is trained based on running the feature binning rule 2, and this model performs feature binning processing on 5 data features, and the names of these 3 data features are age, household register, income, position, and credit respectively. After that, when the server detects that Party A and Party B each run different feature binning rules when establishing the interactive feature binning model 1 and the interactive feature binning model 2 locally, the server then merges the feature binning rule 1 and the feature binning rule 2 to obtain the target feature binning rule, and finally runs the target feature binning rule locally by Party A or Party B to establish the aggregated feature binning model.

[0093] After the server obtains the aggregated feature binning model, it distributes the aggregated feature binning model to different federated participants, so that the corresponding different federated participants can receive the aggregated feature binning model and feedback whether the preset federated training end condition is met based on the aggregated feature binning model. Among them, the preset federated training end condition can be that the model metrics of the aggregated feature binning model reach the preset model metric threshold, and meet indicators such as the preset KS value (Kolmogorov - Smirnov) and AUC (Area Under the Curve) value.

[0094] It should be noted that in this embodiment, the feedback on whether the preset federated training end condition is met can be provided by the main participant, that is, it is provided by the participant with the preset labeled dataset.

[0095] Step S30, if the feedback information indicating that the preset federated training end condition is not met is received, return to the step of receiving the different model information sent by different federated participants after they respectively establish the interactive feature binning models, until the feedback information indicating that the preset federated training end condition is met is received from the corresponding federated participant, so as to obtain the target federated feature binning model.

[0096] When each participating party feeds back that the preset federated training end condition is not met, or when the main participating party (which can be preset) among each participating party feeds back that the preset federated training end condition is not met, that is, when the server receives feedback information that the preset federated training end condition is not met, return to the step of receiving different model information sent by different federated participating parties after respectively establishing an interactive feature binning model, that is, each participating party re-trains the aggregated feature binning model distributed by the server based on the local training data until the corresponding federated participating party feeds back that the preset federated training end condition is met, so as to obtain the target federated feature binning model.

[0097] In this embodiment, it should be noted that the preset federated training end condition may be that the model metrics of the aggregated feature binning model reach the preset model metric threshold, etc.

[0098] A federated modeling method for a feature binning model provided by the present application includes: the server receives different model information sent by different federated participating parties after respectively establishing an interactive feature binning model; combines the different model information to obtain an aggregated feature binning model, and distributes the aggregated feature binning model to the different federated participating parties for the corresponding different federated participating parties to receive the aggregated feature binning model and feedback whether the preset federated training end condition is met based on the aggregated feature binning model; if feedback information that the preset federated training end condition is not met is received, return to the step of receiving different model information sent by different federated participating parties after respectively establishing an interactive feature binning model until feedback information that the corresponding federated participating party feeds back that the preset federated training end condition is met is obtained, so as to obtain the target federated feature binning model. The present application realizes the joint modeling of an interactive feature binning model by multiple federated participating parties, enabling the modeling process of the interactive feature binning model to use more training data on the basis of ensuring data privacy and security, thereby improving the construction efficiency of the feature binning model, ensuring the accuracy and robustness of the model, and ensuring the privacy and security of the modeling data.

[0099] Further, based on the first embodiment of the present application, a second embodiment of a federated modeling method for a feature binning model provided by the present application is proposed. In the second embodiment of a federated modeling method for a feature binning model provided by the present application, in the above step S20, the step of "distributing the aggregated feature binning model to the different federated participating parties for the corresponding different federated participating parties to receive the aggregated feature binning model and feedback whether the preset federated training end condition is met" may include:

[0100] Step A1: Distribute the aggregated feature binning model to the different federal participants. After each corresponding federal participant receives the aggregated feature binning model, calculate the model metrics of the aggregated feature binning model and feedback whether the preset federal training end condition is met. The preset federal training end condition includes reaching a preset model metric threshold or reaching a preset maximum number of federal training iteration rounds.

[0101] It should be noted that in this embodiment, the aggregated feature binning model is distributed to the different federal participants. After each corresponding federal participant receives the aggregated feature binning model, calculate the model metrics locally for the aggregated feature binning model. For example, calculate the woe (weight of Evidence) value and iv (Information Value) value of the aggregated feature binning model, etc. Thus, different federal participants can compare the calculated model metrics with the preset model metric threshold (which can be custom - set) to determine whether the aggregated feature binning model meets the preset federal training end condition and feedback to the server whether the preset federal training end condition is met. And, in the case of judging and feedbacking to the server that the preset federal training end condition is not met, the participant repeats the local iterative training for the current aggregated feature binning model back - transmitted by the server using the training data until the local iterative training reaches the preset maximum number of federal training iteration rounds, and then transmits the model information to the server for model parameter aggregation, and continuously repeats the process of local training and model parameter aggregation until the preset federal training end condition is reached.

[0102] In this embodiment, by receiving different model information sent by different federal participants after establishing interactive feature binning models respectively; where the different federal participants obtain the feature binning model to be operated and the operation task for operating the feature binning model to be operated, and operate the feature binning model to be operated according to the operation task to obtain an interactive feature binning model, and determine model information according to the interactive feature binning. This lays a foundation for accurately obtaining the target federal feature binning model.

[0103] Furthermore, based on the first and second embodiments of the federal modeling method for a feature binning model provided in the present application, a third embodiment of the federal modeling method for a feature binning model provided in the present application is proposed. Please refer to Figure 2, in the third embodiment of the federated modeling method of a feature binning model provided in this application, in the above step S30, if feedback information indicating that the preset federated training end condition is not reached is received, return to the step of receiving different model information sent by different federated participants after establishing interactive feature binning models respectively, until feedback information indicating that the preset federated training end condition is reached is fed back by the corresponding federated participant, so as to obtain the target federated feature binning model. After that, a federated modeling method of a feature binning model provided in this application may further include:

[0104] Step S40, obtain the data to be processed, and input the data to be processed into the target federated feature binning model;

[0105] Step S50, perform binning processing on the data to be processed based on the target federated feature binning model to obtain the data after binning processing, and establish a machine learning model based on the data after binning processing.

[0106] In this embodiment, after the server obtains the target federated feature binning model, if the data to be processed is detected, the data to be processed is obtained and input into the target federated feature binning model. Since the target federated feature binning model is a trained model, therefore, the data to be processed can be subjected to feature binning processing based on the target federated feature binning model to form modeling data specifically used for constructing a machine learning model.

[0107] In this embodiment, by obtaining the data to be processed, inputting the data to be processed into the target federated feature binning model; and establishing a machine learning model based on the data to be processed after binning processing by the target federated feature binning model. Thus, the server can directly perform accurate feature binning processing through the target federated feature binning model to form modeling data, without the need for manual interactive adjustment, avoiding the impact on the privacy and security of data during the data feature binning process, further ensuring the privacy and security of the modeling data, and also improving the construction efficiency of the machine learning model.

[0108] Further, based on the first embodiment and the second embodiment of the federated modeling method of a feature binning model provided in this application, an embodiment of the federated modeling method of a feature binning model provided in this application applied to different federated participants connected to the server is proposed. In the embodiment of this application, a federated modeling method of a feature binning model provided in this application may further include:

[0109] Step A, after establishing an interactive feature binning model, upload the model information of the interactive feature binning model to the server;

[0110] It should be noted that in this embodiment, the application scenario of the federated modeling method for the feature binning model can also be: it is necessary to establish a classification feature binning model with effective prediction and manual intervention. In order to improve the generalization ability of the model, therefore, during the modeling process, the people participating in the modeling are divided into multiple groups, that is, it is through multiple groups of manual intervention. Each group of participants obtains the classification feature binning model corresponding to their group based on the intervention method of their group and the classification data within their group. Then, the classification feature binning models obtained by federating or combining multiple groups are used to obtain the target classification federated feature binning model.

[0111] Specifically, each participant has a classification model dataset: X, where X contains n data items x1, x2, x3,..., xi,..., xn. Each data item xi contains k values (xi1, xi2, xi3, xi4,..., xik). The main participant's corresponding model dataset has a classification label Y: Y contains n data items y1, y2, y3, y4,..., yi, etc. For a classification model (binary classification or multi-classification), yi is discrete.

[0112] Or, the application scenario of the federated modeling method for the feature binning model of this application can also be: it is necessary to establish a regression feature binning model with effective prediction and manual intervention. In order to improve the generalization ability of the model, therefore, during the modeling process, the people participating in the modeling are divided into multiple groups, that is, it is through multiple groups of manual intervention. Each group of participants obtains the regression feature binning model corresponding to their group based on the intervention method of their group and the regression data within their group. Then, the regression feature binning models obtained by federating or combining multiple groups are used to obtain the target federated regression feature binning model.

[0113] Specifically, each participant has a model dataset: X, where X contains n data items x1, x2, x3,..., xi,..., xn. Each data item xi contains k values (xi1, xi2, xi3, xi4,..., xik). In this embodiment, the model dataset has a label Y: Y contains n data items y1, y2, y3, y4,..., yi, etc. For a regression model (binary classification or multi-classification), yi is continuous.

[0114] After different federal participants respectively establish interactive random forest models and send different model information to the server. That is, after different federal participants respectively establish interactive feature binning models through interactive operations, they respectively send the corresponding local model information to the server. Among them, the model information can be information such as model parameters. Specifically, the model parameters can be the feature binning rules run locally when establishing the interactive feature binning model, the feature types of the data features subjected to feature binning processing, and the number of features, etc. Specifically, after Party A establishes the interactive feature binning model 1, it sends the model information 1 corresponding to the interactive random feature binning 1 to the server. After Party B establishes the interactive feature binning model 2, it also sends the model information 2 corresponding to the interactive feature binning model 2 to the server.

[0115] Step B: Receive the aggregated feature binning model obtained by the server by merging the model information, and feedback whether the preset federal training end condition is reached based on the aggregated feature binning model;

[0116] In this embodiment, after the server obtains the different model information of different federal participants, it merges the different model information to obtain an aggregated feature binning model. Then, the server distributes the aggregated feature binning model to different federal participants again. Thus, different federal participants receive the aggregated feature binning model and feedback whether the preset federal training end condition is reached based on the aggregated feature binning model. Among them, the preset federal training end condition can be that the model metrics of the aggregated feature binning model reach the preset model metric threshold, and meet preset indicators such as the KS value (Kolmogorov - Smirnov) and the AUC (Area Under the Curve) value.

[0117] It should be noted that in this embodiment, feedback on whether the preset federal training end condition is reached can be provided by the primary participant, that is, by the participant with the preset labeled dataset.

[0118] In addition, it should be noted that in this embodiment, the server distributes the aggregated feature binning model to the different federal participants so that after the corresponding different federal participants receive the aggregated feature binning model, they calculate the model metrics of the aggregated feature binning model and feedback whether the preset federal training end condition is reached. Among them, the preset federal training end condition includes reaching the preset model metric threshold.

[0119] It should be noted that in this embodiment, the aggregated feature binning model is distributed to the different federal participants, so that after receiving the aggregated feature binning model, the corresponding different federal participants calculate the model metrics locally for the aggregated feature binning model, and then judge whether the aggregated feature binning model meets the preset federal training end condition according to the calculated model metrics, and feedback to the server whether the preset federal training end condition is met. Moreover, in the case of judging and feedbacking to the server that the preset federal training end condition is not met, the participant repeats the local iterative training for the aggregated feature binning model returned by the server this time using the training data locally until the local iterative training reaches the preset local training times, and then transmits the model information to the server for aggregation of model parameters, and continuously repeats the process of local training and model parameter aggregation until the preset federal training end condition is reached.

[0120] Step C, if the feedback information indicating that the preset federal training end condition is not met is sent to the server, return to the step of uploading the model information of the interactive feature binning model to the server after establishing the interactive feature binning model until the feedback information indicating that the preset federal training end condition is met is sent to the server, so that the server can obtain the target federal feature binning model.

[0121] When each participant feedbacks that the preset federal training end condition is not met, or when the main participant (which can be preset) among each participant feedbacks that the preset federal training end condition is not met, that is, when the server receives the feedback information indicating that the preset federal training end condition is not met, return to the step of receiving the different model information sent by different federal participants after establishing the interactive feature binning model respectively. That is, each participant retrains the aggregated feature binning model distributed by the server based on the local training data until the corresponding federal participant feedbacks that the preset federal training end condition is met to obtain the target federal feature binning model.

[0122] In this embodiment, it should be noted that the preset federal training end condition can be that the model metrics of the aggregated feature binning model reach the preset model metric threshold, etc.

[0123] A federated modeling method for a feature binning model provided by this application is as follows: Different federated participants locally establish an interactive feature binning model and upload the model information of the interactive feature binning model to the server; receive the aggregated feature binning model obtained by the server by merging the model information, and based on the aggregated feature binning model, feedback whether the preset federated training end condition is reached; if the feedback information indicating that the preset federated training end condition is not reached is sent to the server, return to the step of establishing the interactive feature binning model and uploading the model information of the interactive feature binning model to the server until the feedback information indicating that the preset federated training end condition is reached is sent to the server, so that the server can obtain the target federated feature binning model. This application realizes the joint modeling of the interactive feature binning model by multiple federated participants, enabling the modeling process of the interactive feature binning model to use more training data on the basis of ensuring data privacy and security, thereby improving the construction efficiency of the feature binning model, ensuring the accuracy and robustness of the model, and ensuring the privacy and security of the modeling data.

[0124] Refer to Figure 3 , Figure 3 It is a schematic diagram of the device structure of the hardware operating environment involved in the solution of the embodiment of this application.

[0125] As Figure 3 shown, the federated modeling device for the feature binning model may include: a processor 1001, such as a CPU, a memory 1005, and a communication bus 1002. Among them, the communication bus 1002 is used to realize the connection and communication between the processor 1001 and the memory 1005. The memory 1005 may be a high-speed RAM memory or a stable memory (non-volatile memory), such as a disk memory. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0126] Optionally, the federated modeling device for the feature binning model may further include a rectangular user interface, a network interface, a camera, an RF (Radio Frequency) circuit, sensors, an audio circuit, a WiFi module, etc. The rectangular user interface may include a display screen and an input sub-module such as a keyboard. Optionally, the rectangular user interface may further include a standard wired interface and a wireless interface. The network interface may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface).

[0127] Those skilled in the art can understand, Figure 3The structure of the federated modeling device of the feature binning model shown does not limit the federated modeling device of the feature binning model, and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0128] As Figure 3 shown, the memory 1005, as a computer storage medium, may include an operating system, a network communication module, and a program for the federated modeling method of the feature binning model. The operating system is a program that manages and controls the hardware and software resources of the federated modeling device of the feature binning model, and supports the operation of the program for the federated modeling method of the feature binning model and other software and / or programs. The network communication module is used to implement communication between the components inside the memory 1005, and communication with other hardware and software in the system of the federated modeling method of the feature binning model.

[0129] In Figure 3 the federated modeling device of the feature binning model shown, the processor 1001 is used to execute the program for the federated modeling method of the feature binning model stored in the memory 1005, and implement the steps of the federated modeling method of the feature binning model described in any one of the above.

[0130] The specific implementation manner of the federated modeling device of the feature binning model in this application is basically the same as the embodiments of the federated modeling method of the feature binning model above, and will not be elaborated here.

[0131] This application embodiment also provides a federated modeling device for a feature binning model, which is applied to a server. The federated interactive modeling device for the random forest model includes:

[0132] A first receiving module, configured to receive different model information sent by different federated participants after respectively establishing an interactive feature binning model;

[0133] A merging and distributing module, configured to merge the different model information to obtain an aggregated feature binning model, and distribute the aggregated feature binning model to the different federated participants, so that the corresponding different federated participants receive the aggregated feature binning model, and based on the aggregated feature binning model, feedback whether a preset federated training end condition is reached;

[0134] A first judging module, configured to, if feedback information indicating that the preset federated training end condition is not reached is received, return to the step of receiving different model information sent by different federated participants after respectively establishing an interactive feature binning model, until feedback information indicating that the preset federated training end condition is reached is received from the corresponding federated participant, so as to obtain a target federated feature binning model;

[0135] Among them, the federated interactive modeling device of the random forest model is also applied to different federated participants connected to the server, and the federated interactive modeling device of the random forest model further includes:

[0136] A model construction module, configured to establish an interactive feature binning model and upload the model information of the interactive feature binning model to the server;

[0137] A second receiving module, configured to receive the aggregated feature binning model obtained by the server merging the model information, and feedback whether a preset federated training end condition is reached based on the aggregated feature binning model;

[0138] A second judgment module, configured to, if feedback information indicating that the preset federated training end condition is not reached is fed back to the server, return to the step of establishing an interactive feature binning model and uploading the model information of the interactive feature binning model to the server until feedback information indicating that the preset federated training end condition is reached is fed back to the server, so that the server obtains a target federated feature binning model.

[0139] Optionally, the merging and distributing module includes:

[0140] A first determination unit, configured to determine, based on the different model information, a first model that performs binning processing on data features and a second model that does not perform binning processing on the data features in the interactive feature binning models of different federated participants;

[0141] Wherein, the data feature is the data feature of the modeling data used by the different federated participants to establish the interactive feature binning model;

[0142] A second determination unit, configured to determine the first model as the aggregated feature binning model.

[0143] Optionally, the merging and distributing module further includes:

[0144] A third determination unit, configured to determine whether the feature binning rules run when different federated participants establish interactive feature binning models are the same based on the different model information;

[0145] A merging unit, configured to merge the feature binning rules to obtain an aggregated feature binning model if the feature binning rules are different.

[0146] Optionally, the merging and distributing module further includes:

[0147] A distribution unit for distributing the aggregated feature binning model to the different federal participants, so that after the corresponding different federal participants receive the aggregated feature binning model, they calculate the model metrics of the aggregated feature binning model and feedback whether the preset federal training end condition is met, where the preset federal training end condition includes reaching a preset model metric threshold or reaching a preset maximum number of federal training iterations.

[0148] Optionally, the first receiving module is further configured to receive different model information sent by different federal participants after respectively establishing an interactive feature binning model, where the different federal participants obtain a feature binning model to be operated and an operation task for operating the feature binning model to be operated, and operate the feature binning model to be operated according to the operation task to obtain an interactive feature binning model, and determine model information according to the interactive feature binning model.

[0149] Optionally, the federal interactive modeling device of the random forest model further includes:

[0150] An acquisition module for acquiring data to be processed and inputting the data to be processed into the target federal feature binning model;

[0151] A binning module for performing binning processing on the data to be processed based on the target federal feature binning model to obtain binned data, and establishing a machine learning model based on the binned data.

[0152] The specific implementation manner of the federal modeling device of the feature binning model in this application is basically the same as that of each embodiment of the above-mentioned federal modeling method of the feature binning model, and will not be elaborated here.

[0153] This application also provides a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the steps of the method for federal modeling of the feature binning model described in any one of the above.

[0154] The specific implementation manner of the computer program product of this application is basically the same as that of each embodiment of the above-mentioned federal modeling method of the feature binning model, and will not be elaborated here.

[0155] The above are only the preferred embodiments of this application, and do not limit the patent scope of this application. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of this application, or directly or indirectly applied to other related technical fields, shall be equally included in the patent scope of this application.

Claims

1. A federated modeling method for a feature binning model, characterized in that Applied to a server, the server is connected to multiple federal participants, and the federated modeling method of the feature binning model includes: Receiving different model information sent by each of the federal participants after establishing an interactive feature binning model respectively. Among them, each of the federal participants locally trains the interactive feature binning model based on the training data stored locally, and each of the federal participants determines the federal model information according to the trained interactive feature binning model, and sends the model information to the server. The model information includes the number of data features and the data feature names for which the interactive feature binning model performs feature binning processing; Merging the different model information to obtain an aggregated feature binning model, and distributing the aggregated feature binning model to the different federal participants for the corresponding different federal participants to receive the aggregated feature binning model and feedback whether the preset federal training end condition is reached based on the aggregated feature binning model; If feedback information indicating that the preset federal training end condition is not reached is received, return to the step of receiving different model information sent by different federal participants after establishing an interactive feature binning model respectively, until feedback information indicating that the preset federal training end condition is reached is received from the corresponding federal participant to obtain a target federal feature binning model; Obtaining data to be processed and inputting the data to be processed into the target federal feature binning model; Performing binning processing on the data to be processed based on the target federal feature binning model to obtain binned data, and establishing a machine learning model based on the binned data.

2. The federated modeling method of the feature binning model according to claim 1, wherein The step of merging the different model information to obtain an aggregated feature binning model includes: Determining, based on the different model information, a first model that performs binning processing on data features and a second model that does not perform binning processing on the data features in the interactive feature binning models of different federal participants; Wherein, the data feature is the data feature of the modeling data used by different federal participants when establishing the interactive feature binning model; Determining the first model as the aggregated feature binning model.

3. The federated modeling method of the feature binning model according to claim 1, characterized in that The step of merging the different model information to obtain an aggregated feature binning model further includes: Determining whether the feature binning rules run when different federal participants establish the interactive feature binning model are the same based on the different model information; If the feature binning rules are different, then merge the feature binning rules to obtain an aggregated feature binning model.

4. The federated modeling method of the feature binning model according to claim 1, characterized in that, The step of distributing the aggregated feature binning model to the different federal participants for the corresponding different federal participants to receive the aggregated feature binning model and feedback whether the preset federal training end condition is reached based on the aggregated feature binning model includes: Distributing the aggregated feature binning model to the different federal participants for the corresponding different federal participants to receive the aggregated feature binning model, calculating the model metrics of the aggregated feature binning model and feedbacking whether the preset federal training end condition is reached, wherein the preset federal training end condition includes reaching a preset model metric threshold or reaching a preset maximum number of federal training iteration rounds.

5. The federated modeling method of the feature binning model according to claim 1, characterized in that The step of receiving different model information sent by different federal participants after respectively establishing an interactive feature binning model includes: Receiving different model information sent by different federal participants after respectively establishing an interactive feature binning model. Among them, the different federal participants obtain a feature binning model to be operated and an operation task for operating the feature binning model to be operated, and operate the feature binning model to be operated according to the operation task to obtain an interactive feature binning model, and determine model information according to the interactive feature binning model.

6. The federated modeling method of the feature binning model according to claim 1, characterized in that, After the step of, if receiving feedback information that does not meet the preset federal training end condition, returning to the step of receiving different model information sent by different federal participants after respectively establishing an interactive feature binning model, until the feedback information indicating that the preset federal training end condition is met is fed back by the corresponding federal participant to obtain a target federal feature binning model, the method includes: Obtaining data to be processed and inputting the data to be processed into the target federal feature binning model; Performing binning processing on the data to be processed based on the target federal feature binning model to obtain binned data, and establishing a machine learning model based on the binned data.

7. A federated modeling method for a feature binning model, characterized in that, Applied to different federal participants in the federal modeling method of the feature binning model as described in any one of claims 1 to 6, the federal modeling method of the feature binning model includes: After establishing an interactive feature binning model, uploading the model information of the interactive feature binning model to the server; Receiving the aggregated feature binning model obtained by the server by merging the model information, and feeding back whether the preset federal training end condition is met based on the aggregated feature binning model; If feeding back feedback information indicating that the preset federal training end condition is not met to the server, returning to the step of uploading the model information of the interactive feature binning model to the server after establishing the interactive feature binning model, until feeding back feedback information indicating that the preset federal training end condition is met to the server for the server to obtain a target federal feature binning model.

8. A federated modeling device for a feature binning model, characterized in that The federal modeling device of the feature binning model includes: a memory, a processor, and a program stored on the memory for implementing the federal modeling method of the feature binning model, The memory is used for storing the program for implementing the federal modeling method of the feature binning model; The processor is used for executing the program for implementing the federal modeling method of the feature binning model to implement the steps of the federal modeling method of the feature binning model as described in any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that, A program for implementing the federal modeling method of the feature binning model is stored on the computer-readable storage medium, and the program for implementing the federal modeling method of the feature binning model is executed by the processor to implement the steps of the federal modeling method of the feature binning model as described in any one of claims 1 to 7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the method for federal modeling of the feature binning model as described in any one of claims 1 to 7.

Images