A method, device, equipment and storage medium for optimizing a solution determination model

Abstract

The embodiment of the present invention discloses an optimization method, device, equipment and storage medium for a solution determination model. The method includes: obtaining the to-be-processed data of at least two participants; predicting the to-be-processed data of the at least two participants based on the pre-trained solution prediction sub-model of at least two participants to obtain a prediction result; determining the target participant among the at least two participants, and sending the prediction result to the target participant. The embodiment of the present invention solves the problem in the prior art that the prediction result can only be sent to the participant who provides the label by determining the target participant who receives the prediction result, thereby reducing the transmission process of the prediction result, avoiding the prediction result from being tampered with during the transmission process, and improving the transmission efficiency and security of the prediction result.

Description

Technical Field

[0001] The embodiments of the present invention relate to computer technology, and in particular to an optimization method, device, equipment and storage medium for a solution determination model. Background Art

[0002] Federated learning technology in the field of data security is becoming more and more popular, and sample-aligned federated learning is also called "vertical federated learning". Each participant in sample-aligned federated learning provides different features of the samples involved in modeling, and each participant ultimately holds a part of the model. In the prediction phase after training, the parts of the joint model held by each participant work together to participate in the prediction process.

[0003] However, only the participant who provides labels during the training phase (Party A) can obtain the prediction results during the prediction phase. The participant who does not provide labels during the training phase (Party B) can only obtain the prediction results by having Party A export the prediction results from the model and manually sending them to Party B. The efficiency of obtaining the prediction results is low, and there is a possibility that Party A may modify the prediction results, which reduces the authenticity of the prediction results. Summary of the invention

[0004] The embodiments of the present invention provide a method, device, equipment and storage medium for optimizing a solution determination model to improve the authenticity of prediction results and the efficiency of sending prediction results, thereby optimizing the solution determination model.

[0005] In a first aspect, an embodiment of the present invention provides a method for optimizing a solution determination model, the method comprising:

[0006] Obtaining data to be processed from at least two parties;

[0007] Based on the pre-trained solution prediction sub-models of at least two participants, predicting the to-be-processed data of the at least two participants to obtain a prediction result;

[0008] A target participant among at least two participants is determined, and the prediction result is sent to the target participant.

[0009] In a second aspect, an embodiment of the present invention further provides an optimization device for a solution determination model, the device comprising:

[0010] A data acquisition module, used to acquire the data to be processed of at least two participants;

[0011] A result prediction module, used to predict the to-be-processed data of at least two participants based on the pre-trained solution prediction sub-models of at least two participants to obtain a prediction result;

[0012] The result sending module is used to determine a target participant among at least two participants and send the prediction result to the target participant.

[0013] In a third aspect, an embodiment of the present invention further provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the optimization method for the scheme determination model as described in any embodiment of the present invention is implemented.

[0014] In a fourth aspect, an embodiment of the present invention further provides a storage medium comprising computer executable instructions, which, when executed by a computer processor, are used to execute the optimization method for the solution determination model as described in any embodiment of the present invention.

[0015] The embodiment of the present invention obtains the solution prediction sub-model of at least two participants through the federated learning technology training, obtains the prediction result by using the trained sub-model, determines the target participant who can receive the prediction result, and sends the prediction result to the participant. The problem in the prior art that the prediction result can only be sent to the participant who provides the label in the training stage is solved, and each participant can obtain the prediction result. After the prediction is completed, the participant who provides the label in the training stage does not need to send the prediction result to other participants, which reduces the data transmission process, improves the data transmission efficiency, and avoids the participant who provides the label in the training stage from tampering with the prediction result during the transmission process, thereby improving the security and authenticity of the prediction results obtained by other participants. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a flowchart of an optimization method for a solution determination model in Embodiment 1 of the present invention;

[0017] Figure 2 It is a flowchart of an optimization method for a solution determination model in Embodiment 2 of the present invention;

[0018] Figure 3 It is a structural block diagram of an optimization device for a scheme determination model in Embodiment 3 of the present invention;

[0019] Figure 4 It is a structural diagram of a computer device in Embodiment 4 of the present invention. DETAILED DESCRIPTION

[0020] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for ease of description, only parts related to the present invention, rather than all structures, are shown in the accompanying drawings.

[0021] Embodiment 1

[0022] Figure 1 This is a flow chart of an optimization method for a solution determination model provided in the first embodiment of the present invention. This embodiment is applicable to the case where a prediction result is obtained by using a solution determination model. This method can be executed by an optimization device for a solution determination model. Figure 1 As shown, the method specifically comprises the following steps:

[0023] Step 110: Obtain data to be processed from at least two participants.

[0024] The federated learning system trains and predicts the solution determination model. The federated learning system can be a sample-aligned federated learning system, also known as a "vertical federated learning" system. The solution determination model can be a model that provides financial compensation solutions to participants. The federated learning system trains the solution prediction sub-models of at least two participants, and the solution prediction sub-models are combined to realize the use of the solution determination model. When using the solution determination model, each participant first sends a model prediction instruction to the local storage facility, loads each locally saved solution prediction sub-model into the federated learning system, and inputs the data to be processed to be predicted. The federated learning system obtains the data to be processed and predicts the solution.

[0025] Step 120: Based on the pre-trained solution prediction sub-model of at least two participants, predict the to-be-processed data of at least two participants to obtain a prediction result.

[0026] Among them, the solution prediction sub-model is first trained for at least two participants, and each participant provides partial features of the training samples, and one of the participants provides labels in addition to partial features of the training samples. Each participant first performs local calculations and encryption on its own side, and then interacts with each other multiple times to exchange intermediate data. Finally, the party that provides the label aggregates the intermediate data of multiple parties to determine the loss of model training. The intermediate data is the intermediate result value of the model during the interactive calculation process. The party that provides the label determines whether to continue training based on the loss value. If the loss value meets the preset loss threshold, the training of the solution prediction sub-model of each participant is completed.

[0027] After the solution prediction sub-model is trained, the data to be processed by each participant is input into the corresponding trained solution prediction sub-model. Each solution prediction sub-model first performs local calculations and encryption on its own data, and then performs multiple intermediate data interactions. In the training phase, the party that provides the label performs the final aggregation calculation of the intermediate data of multiple parties to obtain the prediction result. For example, there are two participants, Party A and Party B. In the training phase, Party A provides part of the features and labels of the training samples, and Party B provides another part of the features of the training samples. The loss value of the training phase is determined by Party A, and in the prediction phase, Party A also performs the final calculation of the prediction result.

[0028] Step 130: Determine a target participant among at least two participants, and send a prediction result to the target participant.

[0029] Among them, after obtaining the prediction results, it is necessary to determine the target participants who can receive the prediction results, so as to send the prediction results to the participants. In the prior art, the prediction results can only be sent to the participants who provide labels during the training phase. For example, the participants are Party A and Party B. During the training phase, Party A provides part of the features and labels of the training samples, and Party B provides another part of the features of the training samples. Then Party A is the target participant, and only Party A can get the prediction results. If Party B wants to get the prediction results, it needs to negotiate with Party A, and Party A will send the prediction results to Party B.

[0030] In this embodiment, optionally, determining a target participant among at least two participants includes: obtaining a contribution threshold in the current incentive mechanism; obtaining contribution values ​​of at least two participants in the model training phase; if the contribution value of a participant is greater than the contribution threshold, then taking the participant as the target participant.

[0031] Specifically, the federated learning system determines the contribution value of each participant to the model training during the model training phase and saves it. The contribution value can indicate the role played by the training samples provided by each participant in the model training process, and can be determined based on data validity or data authenticity. Data validity can indicate the improvement in overall model performance after adding the training sample features provided by the participant, and data authenticity indicates whether the training sample features provided by the participant are honest enough.

[0032] An incentive mechanism is pre-set in the federated learning system, which may include rules for sending prediction results, in which a contribution threshold is set. In the prediction stage, the contribution threshold in the current incentive mechanism and the contribution value of each participant in the model training stage are obtained, and the contribution value of each participant in the model training stage is compared with the contribution threshold. If there is a participant whose contribution value is greater than the contribution threshold, the participant is taken as the target participant, and the participant can obtain the prediction result. The federated learning system instructs the participant who provides the label in the training stage to send the prediction result to the participant who meets the contribution threshold requirement; if there is a participant whose contribution value does not exceed the contribution threshold, and the participant does not provide a label in the training stage, the participant is not the target participant and cannot receive the prediction result. The beneficial effect of such a setting is that by comparing the contribution value, the contribution of each participant in the model training process can be quickly obtained, and the target participant is determined according to the contribution value, which can greatly enhance the enthusiasm of the participant who can only provide features in the training stage, and avoid malicious tampering of the prediction result when it is transmitted between the participants after the prediction is completed, thereby improving the transmission efficiency and authenticity of the prediction result.

[0033] In this embodiment, optionally, obtaining contribution values ​​of at least two participants in the model training stage includes: determining contribution values ​​of at least two participants in the model training stage based on intermediate data of at least two participants in the model training stage.

[0034] Specifically, the contribution value is determined in the model training stage. In the model training stage, each participant interacts with the intermediate data, and the contribution value of each participant to the model training is determined based on the intermediate data. For example, Party A transmits its own calculated data to Party B, and Party B further calculates the data transmitted by Party A based on its own training sample characteristics to obtain intermediate data. According to the intermediate data, the degree of improvement of the training of the model by the training sample characteristics of Party B can be determined, thereby obtaining the contribution value of Party B to the model training. If there is Party C, Party B transmits the intermediate data to Party C to further determine the contribution value of Party C. The beneficial effect of such a setting is that the effect of the training sample characteristics provided by each participant on the model training can be determined. The greater the effect, the higher the contribution value, so as to determine the target participant according to the contribution value and optimize the solution determination model.

[0035] In this embodiment, optionally, the contribution values ​​of at least two participants in the model training stage are determined based on the intermediate data of at least two participants in the model training stage, including: determining and recording the information amount of the characteristic variables of the training samples of at least two participants based on the intermediate data of at least two participants in the model training stage, wherein the information amount represents the predictive ability of the characteristic variables of the training samples; determining the contribution values ​​of at least two participants in the model training stage based on the information amount of the characteristic variables of the training samples of at least two participants.

[0036] Specifically, the contribution value can be determined based on the data validity of the training sample features of each participant. The data validity is the extent to which the overall model performance is improved after adding the features provided by the participants. The IV value (Information Value) can be used as a criterion for judging data validity. The IV value is mainly used to encode the input variables and evaluate the predictive ability. The size of the IV value can be used as the amount of information to represent the strength of the predictive ability of the training sample feature variables. Based on the intermediate data, the IV value can be calculated, and the contribution value of each participant in the model training stage can be determined based on the amount of information of the feature variables of the training samples of at least two participants. The higher the IV value, the stronger the predictive ability of the variable and the higher the contribution value. The WOE (Weight of Evidence) can be determined by the binning method, and then the IV value can be calculated based on the WOE. The WOE and IV values ​​can be calculated using the following formula:

[0037]

[0038]

[0039] Among them, i represents the number of bins, WOE i represents the evidence weight of the current bin, b represents the bad samples in the current bin, g represents the good samples in the current bin, and b total Represents all bad samples, g total Represents all good samples. The beneficial effect of this setting is that by calculating the IV value, the calculation accuracy of the contribution value can be improved, a judgment standard can be provided for the contribution value, the target participants can be accurately determined, and the solution determination model can be optimized.

[0040] The technical solution of this embodiment is to obtain the solution prediction sub-model of at least two participants through federated learning technology training, obtain the prediction result by using the trained sub-model, determine the target participant who can receive the prediction result, and send the prediction result to the participant. It solves the problem in the prior art that the prediction result can only be sent to the participant who provides the label in the training stage, and realizes that each participant can obtain the prediction result. After the prediction is completed, the participant who provides the label in the training stage does not need to send the prediction result to other participants, which reduces the data transmission process, improves the data transmission efficiency, and avoids the participant who provides the label in the training stage from tampering with the solution during the transmission process, improves the security and authenticity of other participants in obtaining the solution, and realizes the optimization of the solution determination model.

[0041] Embodiment 2

[0042] Figure 2The flowchart of the optimization method of a solution determination model provided by the second embodiment of the present invention is further optimized based on the above embodiment. The method can be executed by an optimization device for a solution determination model. Figure 2 As shown, the method specifically comprises the following steps:

[0043] Step 210: Obtain data to be processed from at least two parties.

[0044] Step 220: Based on the pre-trained solution prediction sub-models of at least two participants, predict the to-be-processed data of at least two participants to obtain a prediction result.

[0045] Step 230: Obtain at least two participant identifications.

[0046] Before training the model, each participant is first identified, for example, the participants are identified as A and B. Before determining the target participant, the participant identifications of all participants are first obtained, so as to determine the target participant according to the participant identifications.

[0047] Step 240: compare the at least two participant identifiers with the preset candidate participant identifiers for receiving the prediction results, determine that the participant with the same identifier as the candidate participant among the at least two participants is the target participant, and send the prediction results to the target participant.

[0048] Before using the solution determination model to make predictions, each participant first negotiates to determine the participant who can receive the prediction results, and saves the identifier of the participant who can receive the prediction results after negotiation as a candidate participant identifier.

[0049] After the federated learning system obtains the prediction results, it compares the identifiers of all the participants involved in the prediction with the preset candidate participant identifiers, searches for the participant that is consistent with the candidate participant identifier from all the participant identifiers involved in the prediction, and determines the found participant as the target participant, and the target participant can receive the prediction results. For example, the participant identifiers are A, B, and C, and the candidate participant identifiers that can receive the prediction results after negotiation are A and B, then participants A and B can receive the prediction results, and C cannot receive the prediction results. The beneficial effect of this setting is that the target participant can be directly determined through negotiation without the need for calculation, which improves the efficiency of determining the target participant, and the prediction results can be sent directly to the target participant to prevent the prediction results from being maliciously tampered with, thereby optimizing the solution determination model.

[0050] The embodiment of the present invention obtains the solution prediction sub-model of at least two participants through federated learning technology training, obtains the prediction result by using the trained sub-model, determines the target participant who can receive the prediction result according to the negotiation of the participants, and sends the prediction result directly to the participants. This solves the problem in the prior art that the prediction result can only be sent to the participant who provides the label in the training stage, and realizes that each participant can obtain the prediction result. After the prediction is completed, the participant who provides the label in the training stage does not need to send the prediction result to other participants, which reduces the data transmission process, improves the efficiency of data transmission and the efficiency of determining the target participant, and avoids the participant who provides the label in the training stage from tampering with the solution during the transmission process, improves the security and authenticity of other participants in obtaining the solution, and realizes the optimization of the solution determination model.

[0051] Embodiment 3

[0052] Figure 3 This is a structural block diagram of an optimization device for a solution determination model provided in the third embodiment of the present invention, which can execute the optimization method for the solution determination model provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method. Figure 3 As shown, the device specifically includes:

[0053] The data acquisition module 301 is used to acquire the data to be processed of at least two participants;

[0054] The result prediction module 302 is used to predict the to-be-processed data of at least two participants based on the pre-trained solution prediction sub-models of at least two participants to obtain a prediction result;

[0055] The result sending module 303 is used to determine a target participant among at least two participants and send the prediction result to the target participant.

[0056] Optionally, the result sending module 303 includes:

[0057] A threshold acquisition unit, used to acquire the contribution threshold in the current incentive mechanism;

[0058] A contribution value acquisition unit, used to acquire contribution values ​​of at least two participants in the model training phase;

[0059] The target determination unit is used to take the participant as a target participant if the contribution value of the participant is greater than a contribution threshold.

[0060] Optionally, a contribution value obtaining unit includes:

[0061] The contribution value determination unit is used to determine the contribution values ​​of at least two participants in the model training stage according to the intermediate data of at least two participants in the model training stage.

[0062] Optionally, a contribution value determination unit is specifically used to:

[0063] Determine and record the information amount of the characteristic variables of the training samples of at least two participants based on the intermediate data of at least two participants in the model training phase, wherein the information amount represents the predictive ability of the characteristic variables of the training samples;

[0064] Determine contribution values ​​of at least two participants in the model training phase according to information amounts of characteristic variables of training samples of at least two participants.

[0065] Optionally, the result sending module 303 is further specifically used for:

[0066] Obtain at least two party identifiers;

[0067] The at least two participant identifiers are compared with the preset candidate participant identifiers for receiving the prediction results, and the participant among the at least two participants that is consistent with the candidate participant identifier is determined as the target participant.

[0068] The embodiment of the present invention obtains a solution prediction sub-model of at least two participants through federated learning technology training, obtains a prediction result by using the trained sub-model, determines the target participant who can receive the prediction result, and sends the prediction result to the participant. This solves the problem in the prior art that the prediction result can only be sent to the participant who provides the label in the training stage, and realizes that each participant can obtain the prediction result. After the prediction is completed, the participant who provides the label in the training stage does not need to send the prediction result to other participants, which reduces the data transmission process, improves the data transmission efficiency, and avoids the participant who provides the label in the training stage from tampering with the solution during the transmission process, improves the security and authenticity of other participants in obtaining the solution, and realizes the optimization of the solution determination model.

[0069] Embodiment 4

[0070] Figure 4 It is a structural diagram of a computer device provided in Embodiment 4 of the present invention. Figure 4 A block diagram of an exemplary computer device 400 suitable for use in implementing embodiments of the present invention is shown. Figure 4 The computer device 400 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0071] like Figure 4 As shown, the computer device 400 is in the form of a general computing device. The components of the computer device 400 may include, but are not limited to: one or more processors or processing units 401, a system memory 402, and a bus 403 connecting different system components (including the system memory 402 and the processing unit 401).

[0072] Bus 403 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor or a local bus using any of a variety of bus structures. For example, these architectures include but are not limited to Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MAC) bus, Enhanced ISA bus, Video Electronics Standards Association (VESA) local bus and Peripheral Component Interconnect (PCI) bus.

[0073] Computer device 400 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by computer device 400, including volatile and non-volatile media, removable and non-removable media.

[0074] System memory 402 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 404 and / or cache memory 405. Computer device 400 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 406 may be used to read and write non-removable, non-volatile magnetic media ( Figure 4 not shown, usually called a "hard drive"). Although Figure 4 Not shown in the figure, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk"), and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, a DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 403 via one or more data medium interfaces. Memory 402 may include at least one program product having a set (e.g., at least one) of program modules that are configured to perform the functions of various embodiments of the present invention.

[0075] A program / utility 408 having a set (at least one) of program modules 407 may be stored, for example, in memory 402, such program modules 407 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment. Program modules 407 generally perform the functions and / or methods of the embodiments described herein.

[0076] The computer device 400 may also communicate with one or more external devices 409 (e.g., keyboard, pointing device, display 410, etc.), may also communicate with one or more devices that enable a user to interact with the computer device 400, and / or communicate with any device that enables the computer device 400 to communicate with one or more other computing devices (e.g., network card, modem, etc.). Such communication may be performed via input / output (I / O) interface 411. Furthermore, the computer device 400 may also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN) and / or public network, such as the Internet) via network adapter 412. As shown, the network adapter 412 communicates with other modules of the computer device 400 via bus 403. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the computer device 400, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0077] The processing unit 401 executes various functional applications and data processing by running the program stored in the system memory 402, for example, the optimization method of the solution determination model provided in the embodiment of the present invention includes:

[0078] Obtaining data to be processed from at least two parties;

[0079] Based on the pre-trained solution prediction sub-models of at least two participants, predicting the to-be-processed data of at least two participants to obtain a prediction result;

[0080] A target participant among at least two participants is determined, and a prediction result is sent to the target participant.

[0081] Embodiment 5

[0082] Embodiment 5 of the present invention further provides a storage medium containing computer executable instructions, on which a computer program is stored. When the program is executed by a processor, the optimization method of the solution determination model provided in the embodiment of the present invention is implemented, including:

[0083] Obtaining data to be processed from at least two parties;

[0084] Based on the pre-trained solution prediction sub-models of at least two participants, predicting the to-be-processed data of at least two participants to obtain a prediction result;

[0085] A target participant among at least two participants is determined, and a prediction result is sent to the target participant.

[0086] The computer storage medium of the embodiment of the present invention can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, 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), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.

[0087] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0088] The program code embodied on the computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0089] Computer program code for performing the operations of the present invention may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0090] Note that the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A method for optimizing a solution determination model, characterized in that: include: Obtaining data to be processed from at least two parties; Based on the pre-trained solution prediction sub-models of at least two participants, predicting the to-be-processed data of the at least two participants to obtain a prediction result; wherein the at least two participants include: a participant who does not provide a label and a participant who provides a label; Determine a target participant among at least two participants, and send the prediction result to the target participant; The solution prediction sub-models of the at least two participants are trained in the following manner: Each participant performs local calculations and encryption on their own side, interacts with each other multiple times, and exchanges intermediate data; The party providing the labels aggregates the intermediate data of multiple parties to determine the loss of model training and determines whether to continue training based on the loss value; If the loss value meets the preset loss value, the training of the solution prediction sub-model of each participant is completed; The method of predicting the at least two participants' solution prediction sub-models based on the pre-trained solution prediction of the at least two participants to obtain the prediction results includes: Input the data to be processed of each participant into the corresponding trained solution prediction sub-model respectively; Each prediction sub-model of each solution performs local calculations and encryption on its own data, and exchanges intermediate data with each other multiple times; The participants who provide labels in the training phase perform the final aggregation calculation of the intermediate data of multiple parties to obtain the prediction results; Wherein, determining the target participant among the at least two participants includes: Get the contribution threshold in the current incentive mechanism; Obtain the contribution values ​​of at least two participants in the model training phase; If the contribution value of the participant is greater than the contribution threshold, the participant is taken as the target participant; The step of obtaining contribution values ​​of at least two participants in the model training phase includes: Determining contribution values ​​of the at least two participants in the model training phase according to intermediate data of the at least two participants in the model training phase; The step of determining the contribution values ​​of the at least two participants in the model training stage according to the intermediate data of the at least two participants in the model training stage includes: Determine and record the information amount of the characteristic variables of the training samples of the at least two participants according to the intermediate data of the at least two participants in the model training phase, wherein the information amount represents the predictive ability of the characteristic variables of the training samples; Determine the contribution values ​​of the at least two participants in the model training stage according to the amount of information of the characteristic variables of the training samples of the at least two participants; wherein the contribution value is determined according to the data validity of the characteristics of the training samples of each participant, the data validity is the extent of improvement in the overall model performance after adding the characteristics provided by the participants, the judgment standard of the data validity is the information value IV value, the IV value is used to evaluate the encoding and prediction capabilities of the input training sample characteristic variables, and the IV value is determined according to the weight of evidence WOE value; the WOE value is determined by a binning method.

2. The method according to claim 1, characterized in that Identifying a target party from among at least two parties, further comprising: Obtain at least two party identifiers; The at least two participant identifiers are compared with a preset candidate participant identifier for receiving the prediction result, and the participant among the at least two participants that is consistent with the candidate participant identifier is determined as the target participant.

3. An optimization device for a solution determination model, characterized in that: include: A data acquisition module, used to acquire the data to be processed of at least two participants; A result prediction module, used for predicting the to-be-processed data of at least two participants based on the pre-trained solution prediction sub-models of the at least two participants to obtain a prediction result; wherein the at least two participants include: a participant who does not provide a label and a participant who provides a label; A result sending module, used for determining a target participant among at least two participants, and sending the prediction result to the target participant; Wherein, the result sending module includes: A threshold acquisition unit, used to acquire the contribution threshold in the current incentive mechanism; A contribution value acquisition unit, used to acquire contribution values ​​of at least two participants in the model training phase; a target determination unit, configured to take the participant as the target participant if the contribution value of the participant is greater than the contribution threshold; Wherein, the contribution value acquisition unit includes: A contribution value determination unit, used to determine the contribution values ​​of at least two participants in the model training stage according to the intermediate data of at least two participants in the model training stage; Wherein, the contribution value determination unit is specifically used to determine and record the information amount of the characteristic variables of the training samples of the at least two participants according to the intermediate data of the at least two participants in the model training stage, wherein the information amount represents the prediction ability of the characteristic variables of the training samples; determine the contribution value of the at least two participants in the model training stage according to the information amount of the characteristic variables of the training samples of the at least two participants; wherein the contribution value is determined according to the data validity of the training sample characteristics of each participant, the data validity is the improvement of the overall model performance after adding the characteristics provided by the participants, and the judgment standard of the data validity is the information value IV value, the IV value is used to evaluate the encoding and prediction ability of the input training sample characteristic variables, and the IV value is determined according to the weight of evidence WOE value; the WOE value is determined by the binning method; The result prediction module is specifically used to input the to-be-processed data of each participant into the corresponding trained solution prediction sub-model; each solution prediction sub-model performs local calculations and encryption on its own data, and performs multiple intermediate data exchanges with each other; the participant who provides labels in the training phase performs the final aggregation calculation of the intermediate data of multiple parties to obtain the prediction result; Among them, the solution prediction sub-models of at least two participants are trained in the following manner: each participant performs local calculations and encryption on its own side, interacts with each other multiple times, and exchanges intermediate data; the participant who provides labels aggregates the intermediate data of multiple parties to judge the loss of model training, and determines whether to continue training based on the loss value; if the loss value meets the preset loss value, the training of the solution prediction sub-models of each participant is completed.

4. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the optimization method of the solution determination model as described in any one of claims 1-2 is implemented.

5. A storage medium containing computer executable instructions, characterized in that: When the computer executable instructions are executed by a computer processor, they are used to perform the optimization method of the solution determination model as described in any one of claims 1-2.

Images