Data processing method and device

Abstract

The embodiments of the present specification provide a data processing method and an apparatus, wherein the data processing method comprises: obtaining parameter configuration information of a target object, sampling parameter values ​​of the object parameters based on a parameter value range corresponding to the object parameters in the parameter configuration information, generating multiple object parameter sets according to the sampling results, wherein each object parameter set includes a parameter value of an object parameter, or includes a combination of parameter values ​​corresponding to at least two object parameters, respectively, performing regression prediction on the target object based on the multiple object parameter sets, determining a target object parameter set from the multiple object parameter sets according to the prediction results, and performing parameter adjustment on the target object based on the target object parameter set.

Description

Technical Field

[0001] The embodiments of the present specification relate to the field of computer technology, and in particular, to a data processing method. Background Art

[0002] There is a need for parameter tuning in scenarios such as operating system kernels, compilers, and databases. More effective parameter settings can optimize service performance and improve overall service quality.

[0003] At present, more and more parameter tuning processes are implemented in an automated way, with the aim of finding the optimal parameters in a reasonable search space in a shorter time. Specifically, the parameters that affect the performance indicators of the target object are adjusted to minimize the adjusted loss function, thereby improving the overall performance of the target object. However, in the actual automated tuning process, the value of the loss function can only be obtained after the parameters are actually applied and the evaluation program (benchmark) is executed. The parameter application and evaluation program execution process are often costly. At the same time, the number of parameters that need to be tuned is often large, and high-dimensional parameters pose a greater challenge to the tuning algorithm, which may not only lead to low tuning efficiency, but also affect the tuning effect. Summary of the invention

[0004] In view of this, an embodiment of this specification provides a data processing method. One or more embodiments of this specification also relate to a data processing device, a computing device, a computer-readable storage medium and a computer program to solve the technical defects existing in the prior art.

[0005] According to a first aspect of an embodiment of this specification, a data processing method is provided, including:

[0006] Get the parameter configuration information of the target object;

[0007] Based on the parameter value range corresponding to the object parameter in the parameter configuration information, sampling the parameter value of the object parameter;

[0008] Generate multiple object parameter sets according to the sampling results, wherein each object parameter set includes a parameter value of an object parameter, or includes a combination of parameter values ​​corresponding to at least two object parameters respectively;

[0009] Regression prediction is performed on the target object based on the multiple object parameter sets, a target object parameter set is determined from the multiple object parameter sets according to the prediction result, and parameters of the target object are adjusted based on the target object parameter set.

[0010] Optionally, determining a target object parameter set from the multiple object parameter sets according to the prediction result includes:

[0011] Using the multiple object parameter sets as training samples and the prediction results as labels, training the score prediction model to be trained to obtain the score prediction model;

[0012] A target object parameter set of the target object is determined according to the multiple object parameter sets and the score prediction model.

[0013] Optionally, determining the target object parameter set according to the multiple object parameter sets and the prediction model includes:

[0014] Generate a to-be-processed object parameter set corresponding to the target object according to a preset algorithm and the plurality of object parameter sets;

[0015] Inputting the parameter set of the object to be processed into the score prediction model for processing, and generating a score prediction result corresponding to each parameter set of the object to be processed;

[0016] The target object parameter set is determined from the to-be-processed object parameter set according to the score prediction result.

[0017] Optionally, generating a to-be-processed object parameter set corresponding to the target object according to a preset algorithm and the plurality of object parameter sets includes:

[0018] Sorting the plurality of object parameter sets according to the prediction results, and setting weights for the object parameters according to the sorting results;

[0019] Determining the object parameters to be adjusted from the object parameters according to the weight setting result;

[0020] According to the preset algorithm and the parameters of the object to be adjusted, a set of parameters of the object to be processed corresponding to the target object is generated.

[0021] Optionally, after obtaining parameter configuration information of the target object, the method further includes:

[0022] A search space is initialized based on parameter configuration information of a target object, wherein the search space includes a parameter value range of an object parameter of the target object.

[0023] Optionally, the sampling the parameter value of the object parameter based on the parameter value range corresponding to the object parameter in the parameter configuration information includes:

[0024] The sampling module samples from the search space to obtain a parameter value sampling result corresponding to the object parameter.

[0025] Optionally, generating a to-be-processed object parameter set corresponding to the target object according to a preset algorithm and the plurality of object parameter sets includes:

[0026] Calling a screening function in a screening module, sorting the plurality of object parameter sets according to the prediction results, and setting weights for the object parameters according to the sorting results;

[0027] Determining the object parameters to be adjusted from the object parameters according to the weight setting result;

[0028] The object parameter generation algorithm in the search strategy module is called to process the object parameters to be adjusted, and generate a set of object parameters to be processed corresponding to the target object.

[0029] Optionally, after determining the target object parameter set from the to-be-processed object parameter set according to the score prediction result, the method further includes:

[0030] Determine the budget allocation value corresponding to the target object parameter set by a budget control module;

[0031] Allocating a budget to the target object parameter set by using a budget allocation function in the budget control module;

[0032] Based on the budget allocation result, an evaluation program is called to process the target object parameter set to generate a score prediction result corresponding to the target object parameter set.

[0033] Optionally, after generating the score prediction result corresponding to the target object parameter set, the method further includes:

[0034] Invoking the object parameter generation algorithm in the search strategy module to process the object parameters to be adjusted in the target object parameter set to generate a set of object parameters to be evaluated corresponding to the target object;

[0035] The to-be-evaluated object parameter set is used as the to-be-processed object parameter set, and the step of determining the budget allocation value corresponding to the target object parameter set by the budget control module is returned to be executed.

[0036] Optionally, the data processing method further includes:

[0037] determining whether a stop processing condition is met based on the processing information of the evaluation program;

[0038] If so, a set of object parameters for adjusting parameters of the target object is determined according to the score prediction result.

[0039] According to a second aspect of an embodiment of this specification, there is provided a data processing device, including:

[0040] An acquisition module, configured to acquire parameter configuration information of a target object;

[0041] a sampling module configured to sample a parameter value of the object parameter based on a parameter value range corresponding to the object parameter in the parameter configuration information;

[0042] A generating module is configured to generate a plurality of object parameter sets according to the sampling results, wherein each object parameter set includes a parameter value of an object parameter, or includes a combination of parameter values ​​corresponding to at least two object parameters respectively;

[0043] The adjustment module is configured to perform regression prediction on the target object based on the multiple object parameter sets, determine a target object parameter set from the multiple object parameter sets according to the prediction result, and perform parameter adjustment on the target object based on the target object parameter set.

[0044] According to a third aspect of an embodiment of this specification, a computing device is provided, including:

[0045] Memory and processor;

[0046] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement any step of the data processing method.

[0047] According to a fourth aspect of the embodiments of this specification, a computer-readable storage medium is provided, which stores computer-executable instructions, and when the instructions are executed by a processor, the steps of any one of the data processing methods are implemented.

[0048] According to a fifth aspect of the embodiments of this specification, a computer program is provided, wherein when the computer program is executed in a computer, the computer is caused to execute the steps of the above-mentioned data processing method.

[0049] An embodiment of the present specification obtains parameter configuration information of a target object, samples parameter values ​​of the object parameters based on a parameter value range corresponding to the object parameters in the parameter configuration information, generates multiple object parameter sets according to the sampling results, wherein each object parameter set includes a parameter value of an object parameter, or a combination of parameter values ​​corresponding to at least two object parameters, performs regression prediction on the target object based on the multiple object parameter sets, determines a target object parameter set from the multiple object parameter sets according to the prediction results, and performs parameter adjustment on the target object based on the target object parameter set.

[0050] The embodiments of this specification disassemble the parameter adjustment process of the target object so that the parameter adjustment algorithm development meets the principles of high cohesion and low coupling, which is beneficial to improving the maintainability and scalability of the parameter adjustment algorithm, and is also beneficial to reducing the difficulty of secondary development; in addition, by sampling the parameter values ​​of each object parameter and generating multiple object parameter sets based on the sampling results, and then selecting a target object parameter set from the multiple object parameter sets to adjust the parameters of the target object based on the target object parameter set, it is beneficial to improve the performance of the parameter adjustment algorithm, thereby helping to improve the accuracy of the parameter adjustment results of the target object. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is a flow chart of a data processing method provided by an embodiment of this specification;

[0052] Figure 2a It is a schematic diagram of a framework provided by an embodiment of this specification;

[0053] Figure 2b is a schematic diagram of a data processing process provided by an embodiment of this specification;

[0054] Figure 3 is a processing flow chart of a data processing method provided by an embodiment of this specification;

[0055] Figure 4 is a structural schematic diagram of a data processing device provided by an embodiment of this specification;

[0056] Figure 5 It is a structural block diagram of a computing device provided by an embodiment of this specification. DETAILED DESCRIPTION

[0057] Many specific details are described in the following description to facilitate a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of this specification, so this specification is not limited to the specific implementation disclosed below.

[0058] The terms used in one or more embodiments of this specification are only for the purpose of describing specific embodiments, and are not intended to limit one or more embodiments of this specification. The singular forms of "a", "said" and "the" used in one or more embodiments of this specification and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0059] It should be understood that although the terms first, second, etc. may be used to describe various information in one or more embodiments of this specification, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of one or more embodiments of this specification, the first may also be referred to as the second, and similarly, the second may also be referred to as the first. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0060] First, the terms involved in one or more embodiments of this specification are explained.

[0061] Parameter tuning: The process of adjusting system parameters to continuously improve the system's score on a benchmark program.

[0062] Parameter tuning algorithm: A program that implements a certain algorithm strategy and continuously gives system parameter combinations to find parameter configurations that improve application performance.

[0063] Algorithm framework: A unified modular implementation solution for a certain algorithm or a certain type of algorithm, which defines the functions of each module and the interaction between modules.

[0064] There is a need for parameter tuning in scenarios such as operating system kernels, compilers, and databases. Effective parameter settings can optimize service performance and improve overall service quality. However, the above fields involve a large number of parameters, and the parameter space dimension is high, which may be as high as hundreds or thousands of dimensions. At the same time, traditional tuning methods often rely on experienced engineers to perform manual tuning based on domain knowledge. High-dimensional parameters pose a great challenge to manual tuning, which may not only lead to low tuning efficiency, but also affect the tuning effect. Therefore, in recent years, academia and industry have relied on machine learning methods to promote the development of algorithms and tools for parameter tuning in scenarios such as operating system kernels, compilers, and databases based on a large amount of tuning data records.

[0065] The current mainstream tuning algorithm is based on Bayesian theory, which dynamically searches in the preset parameter space according to the tuning effect of real-time feedback. This algorithm framework is relatively general. In recent years, it has not only been used in parameter tuning scenarios related to computer systems, but also widely used in hyperparameter tuning scenarios of deep learning models. When there are many parameters and the dimension of the parameter space is high, the search cycle required by the Bayesian algorithm is often long. In order to improve the overall tuning efficiency, some parameter tuning algorithms designed based on the idea of ​​genetic algorithms have also emerged in recent years. These algorithms select parameter configurations that are known to perform well, mutate them, and find better parameter configurations through iteration.

[0066] Based on this, an embodiment of the present specification obtains parameter configuration information of the target object, samples the parameter values ​​of the object parameters based on the parameter value range corresponding to the object parameters in the parameter configuration information, generates multiple object parameter sets according to the sampling results, wherein each object parameter set includes the object parameters, performs regression prediction on the target object based on the multiple object parameter sets, determines the target object parameter set from the multiple object parameter sets according to the prediction results, and adjusts the parameters of the target object based on the target object parameter set.

[0067] The embodiments of this specification disassemble the parameter adjustment process of the target object so that the parameter adjustment algorithm development meets the principles of high cohesion and low coupling, which is beneficial to improving the maintainability and scalability of the parameter adjustment algorithm, and is also beneficial to reducing the difficulty of secondary development; in addition, by sampling the parameter values ​​of each object parameter and generating multiple object parameter sets based on the sampling results, and then selecting a target object parameter set from the multiple object parameter sets to adjust the parameters of the target object based on the target object parameter set, it is beneficial to improve the performance of the parameter adjustment algorithm, thereby helping to improve the accuracy of the parameter adjustment results of the target object.

[0068] In this specification, a data processing method is provided. This specification also relates to a data processing apparatus, a computing device, a computer-readable storage medium, and a computer program, which are described in detail one by one in the following embodiments.

[0069] Figure 1 A flow chart of a data processing method provided according to an embodiment of the present specification is shown, which specifically includes the following steps.

[0070] Step 102: Obtain parameter configuration information of the target object.

[0071] Specifically, the target object may be an operating system kernel, a compiler, or a database, etc.

[0072] Since there are requirements for parameter tuning in scenarios such as the operating system kernel, compiler, and database, more effective parameter settings can optimize service performance and improve overall service quality.

[0073] The data processing method provided in the embodiments of this specification can be used to tune the parameters of the target object, and specifically provides a unified framework for intelligent parameter tuning, including 6 main modules (base classes), namely, search space (SearchSpace), sampling module (SampleFunc), proxy module (Surrogate), search strategy module (Strategy), screening module (Filter) and budget management module (Banker). The specific framework diagram is shown in FIG. Figure 2ashown.

[0074] The purpose of parameter tuning is to find the better parameters of the target object in a reasonable search space in a shorter time. Therefore, when the parameters of the target object need to be adjusted, the parameter configuration information of the target object can be obtained first, and the parameter configuration information can include one or at least two object parameters of the target object. After obtaining the parameter configuration information of the target object, the search space can be initialized based on the parameter configuration information of the target object, wherein the search space includes the parameter value range of each object parameter of the target object.

[0075] Specifically, the search space defines the parameter tuning range of each object parameter in the target object, including the basic data such as the parameter range, parameter value domain, data type, etc., and also includes the method implementation of parameter mutation. Among them, the framework designed in the embodiment of this specification currently defines three parameter types: options, sequence and variable. Options represent discrete parameters, variables represent continuous parameters, and sequences represent discrete parameters, but there is a relative size relationship between parameter values.

[0076] In practical applications, the first stage of parameter tuning for the target object requires parameter configuration sampling and parameter sensitivity analysis. Therefore, the search space can be initialized according to the Knobs configuration file of the target object. In addition, since the parameter tuning range of each object parameter in the target object is defined in the search space, the object parameters of the target object can be sampled based on the parameter tuning range to adjust the parameters of the target object according to the sampling results.

[0077] Step 104: sampling the parameter value of the object parameter based on the parameter value range corresponding to the object parameter in the parameter configuration information.

[0078] In specific implementation, since the framework designed in the embodiment of this specification includes a sampling module, the parameter value of the object parameter is sampled based on the parameter value range corresponding to the object parameter in the parameter configuration information. Specifically, the sampling module is used to sample from the search space to obtain the parameter value sampling result corresponding to the object parameter.

[0079] Specifically, the sampling module (SampleFunc) defines the parameter sampling method, such as random sampling (Random) and Latin hypercube sampling (Latin hypercube sampling), which can be used to sample the initial data. Therefore, after initializing the search space, the sampling method (Sampling method) of the sampling module (SampleFunc object) can be called to perform parameter configuration sampling, so as to obtain the parameter value sampling result corresponding to the object parameter from the search space through the sampling module.

[0080] Take the three object parameters of the target object included in the parameter configuration information as an example. These three object parameters are object parameter 1, object parameter 2 and object parameter 3. In addition, the parameter types of object parameter 1, object parameter 2 and object parameter 3 defined in the search space are all discrete, and the parameter value ranges of object parameter 1, object parameter 2 and object parameter 3 are all [1, 2, 3, 4, 5]. In this case, the sampling method in the sampling module is called to sample the parameter value of each object parameter based on the parameter type and parameter value range of each object parameter in the search space to obtain the three parameter value sampling results corresponding to each object parameter. Among them, the parameter value sampling results of object parameter 1 can be 1, 2, 3; the parameter value sampling results of object parameter 2 are 2, 3, 4; the parameter value sampling results of object parameter 3 are 3, 4, 5.

[0081] Step 106: Generate multiple object parameter sets according to the sampling results, wherein each object parameter set includes a parameter value of an object parameter, or a combination of parameter values ​​corresponding to at least two object parameters.

[0082] Specifically, based on the parameter value range corresponding to the object parameter, after sampling the parameter value of the object parameter to obtain the corresponding sampling result, multiple object parameter sets can be generated according to the sampling result, wherein, when only one object parameter is adjusted, each object parameter set contains a parameter value of the object parameter; or, when two or more object parameters are adjusted, the sampling results of each object parameter can be combined, and each generated object parameter set contains a combination of parameter values ​​corresponding to at least two object parameters.

[0083] Continuing with the above example, parameter value sampling is performed on three object parameters, and each object parameter corresponds to three parameter value sampling results. Therefore, the three object parameters are combined based on the sampling results, and the combination method is: the nine parameter value sampling results of the three object parameters are arranged and combined, and each generated object parameter set contains one object parameter 1, one object parameter 2 and one object parameter 3, and the parameter value combinations of each object parameter in different object parameter sets are different, that is, the parameter values ​​of object parameter 1, object parameter 2 and object parameter 3 in different object parameter sets are not completely equal.

[0084] Step 108: perform regression prediction on the target object based on the multiple object parameter sets, determine a target object parameter set from the multiple object parameter sets according to the prediction result, and adjust parameters of the target object based on the target object parameter set.

[0085] Specifically, after generating multiple object parameter sets, each object parameter set will execute an evaluation program (benchmark) in turn. The evaluation program is used to perform performance evaluation on the target object, that is, to determine the performance of the target object when the parameter values ​​of the object parameters of the target object are consistent with the parameter values ​​of each object parameter in any object parameter set.

[0086] Therefore, the embodiments of the present specification can perform regression prediction on the target object by executing an evaluation procedure on each object parameter set, generate multiple prediction results (benchmark scores) for the target object, and then determine the target object parameter set based on the multiple prediction results, and adjust the parameters of the target object based on the target object parameter set.

[0087] In a specific implementation, determining a target object parameter set from the multiple object parameter sets according to the prediction result includes:

[0088] Using the multiple object parameter sets as training samples and the prediction results as labels, training the score prediction model to be trained to obtain the score prediction model;

[0089] A target object parameter set of the target object is determined according to the multiple object parameter sets and the score prediction model.

[0090] Specifically, the framework designed in the embodiment of this specification also includes a surrogate module, which defines the implementation method of the score prediction model. The score prediction model is a core component of some parameter intelligent tuning algorithms. It models the relationship between the object parameter set and the benchmark score, and predicts the benchmark score by inputting the object parameter set, thereby guiding the parameter intelligent tuning process.

[0091] In addition, the two main functions of the agent module are Fit and Predict, which are responsible for training the model and predicting the benchmark score respectively. Most of the definitions of Surrogate rely on the model (Model) and the loss function (LossFunc) to implement, while some Surrogate (such as RBF) do not need to rely on the loss function, that is, they can be implemented without the participation of the loss function.

[0092] Therefore, after generating multiple object parameter sets and obtaining the regression prediction results corresponding to each object parameter set, the Fit function in the proxy module can be called, and the multiple object parameter sets can be used as training samples. The benchmark scores corresponding to each object parameter set can be used as labels to train the score prediction model to be trained to obtain the score prediction model, and determine the target object parameter set of the target object based on the multiple object parameter sets and the score prediction model.

[0093] Among them, during the model training process, each object parameter set can be processed by the score prediction model to be trained to generate a prediction result corresponding to each object parameter set, and then the loss value is calculated by combining the prediction result and the benchmark score corresponding to each object parameter set. Then, the model is iteratively trained according to the loss value, that is, the model parameters are adjusted until the calculated loss value is less than the preset threshold, or the number of iterative training times is greater than or equal to the preset threshold. The model training process can be stopped to obtain the trained score prediction model.

[0094] Further, determining the target object parameter set according to the multiple object parameter sets and the score prediction model includes:

[0095] Generate a to-be-processed object parameter set corresponding to the target object according to a preset algorithm and the plurality of object parameter sets;

[0096] Inputting the parameter set of the object to be processed into the score prediction model for processing, and generating a score prediction result corresponding to each parameter set of the object to be processed;

[0097] The target object parameter set is determined from the to-be-processed object parameter set according to the score prediction result.

[0098] Specifically, since the score prediction model is used to model the relationship between the object parameter set and the benchmark score, the benchmark score is predicted by inputting the object parameter set, thereby guiding the parameter intelligent tuning process.

[0099] Therefore, after training and obtaining the score prediction model, a new set of object parameters to be processed can be generated based on multiple object parameter sets, and the set of parameter parameters to be processed can be input into the score prediction model so that the score prediction model can determine the performance of the target object when the parameter values ​​of the object parameters of the target object are consistent with the parameter values ​​of each object parameter in any set of object parameters to be processed.

[0100] Among them, the Predict function in the proxy module can be called, and multiple object parameter sets can be processed using a preset algorithm to generate a set of object parameter sets to be processed corresponding to the target object, and then the set of object parameter sets to be processed can be input into the score prediction model for processing to generate a score prediction result corresponding to each set of object parameter sets to be processed.

[0101] In practical applications, the preset algorithm may be a genetic mutation algorithm.

[0102] In a specific implementation, according to a preset algorithm and the plurality of object parameter sets, a set of object parameters to be processed corresponding to the target object is generated, including:

[0103] Sorting the plurality of object parameter sets according to the prediction results, and setting weights for the at least two object parameters according to the sorting results;

[0104] Determining the object parameter to be adjusted from the at least two object parameters according to the weight setting result;

[0105] According to the preset algorithm and the parameters of the object to be adjusted, a set of parameters of the object to be processed corresponding to the target object is generated.

[0106] Furthermore, according to the preset algorithm and the plurality of object parameter sets, a set of object parameters to be processed corresponding to the target object is generated, including:

[0107] Calling a screening function in a screening module, sorting the plurality of object parameter sets according to the prediction result, and setting weights for the at least two object parameters according to the sorting result;

[0108] Determining the object parameter to be adjusted from the at least two object parameters according to the weight setting result;

[0109] The object parameter generation algorithm in the search strategy module is called to process the object parameters to be adjusted, and generate a set of object parameters to be processed corresponding to the target object.

[0110] Specifically, the framework designed in the embodiments of this specification also includes a screening module and a search strategy module.

[0111] Among them, the filter module (Filter) is used to select which object parameters in the search space are more worthy of parameter adjustment and which object parameters have little impact on the performance of the target object. The main function of the filter module is the allocation function (Allocate), which is used to select and assign weights to object parameters in the search space. Filter involves the implementation of the sensitive parameter analysis algorithm and is an optional component.

[0112] Therefore, after performing regression prediction on each object parameter set and obtaining the corresponding prediction results, the grant function of the screening module can be called according to the parameter value of each object parameter in each object parameter set and the prediction result of the object parameter set to sort the multiple object parameter sets, and set the weight of each object parameter according to the sorting result. Among them, each object parameter set can be sorted in descending order according to the size of the prediction result, and the weight of each object parameter can be set according to the parameter value of each object parameter in the first n object parameter sets in the sorting result.

[0113] In practical applications, n is a positive integer, and the value of n can be determined according to actual needs. In addition, when setting the weight of the target object parameter, the change range of the parameter value of the target object parameter in the first n object parameter sets can be referred to. If the change range of the target object parameter in the first n object parameter sets is relatively small relative to the value range of the target object parameter, it indicates that the target object parameter has little impact on the performance of the target object, and the weight of the target object parameter can be set to a smaller value.

[0114] After setting weights for each object parameter, the object parameter to be adjusted can be selected according to the weight of each object parameter. In practical applications, the object parameter with a larger weight can be used as the object parameter to be adjusted.

[0115] In addition, the search strategy module (Strategy) defines the algorithm's methods for generating and screening objects. The corresponding two main functions are the object parameter generation function (Generate) and the evaluation function (Evaluate), which together constitute the backbone of the parameter tuning algorithm, that is, how to generate some object parameters, and then how to screen some object parameters from the generated object parameters to execute the evaluation procedure.

[0116] Therefore, after determining the object parameters to be adjusted from at least two object parameters according to the weight setting result, the object parameter generation algorithm in the search strategy module can be called to process the object parameters to be adjusted, generate new parameter values ​​of the object parameters to be adjusted, and combine the object parameters to be adjusted based on the new parameter values ​​to generate a set of object parameters to be processed of the target object.

[0117] In addition, after the target object parameter set is determined from the to-be-processed object parameter set according to the score prediction result, the budget allocation value corresponding to the target object parameter set may be determined by the budget control module;

[0118] Allocating a budget to the target object parameter set by using a budget allocation function in the budget control module;

[0119] Based on the budget allocation result, an evaluation program is called to process the target object parameter set to generate a score prediction result corresponding to the target object parameter set.

[0120] Specifically, after generating multiple object parameter sets, each object parameter set will execute an evaluation program (benchmark) in turn. The evaluation program is used to perform performance evaluation on the target object, that is, to determine the performance of the target object when the parameter values ​​of the object parameters of the target object are consistent with the parameter values ​​of each object parameter in any object parameter set.

[0121] Specifically, the framework designed in the embodiment of this specification also includes a budget control module (Banker), which is used to determine how much budget each set of object parameters to be processed has to execute the evaluation procedure. It involves the implementation of the multi-fidelity algorithm and is an optional component.

[0122] Therefore, after generating the set of object parameters to be processed, the budget control module can be used to determine the budget allocation value corresponding to each set of object parameters to be processed, and the budget allocation function (Allocate function) in the budget control module can be used to allocate a budget to each set of object parameters to be processed.

[0123] After allocating a budget to each set of object parameters to be processed, each set of object parameters to be processed needs to be processed by an evaluation program to perform regression prediction on the target object, thereby obtaining a corresponding prediction result. However, since the budget can be used to characterize the processing time of the evaluation program for the set of object parameters to be processed, the longer the processing time of the evaluation program, the more accurate the prediction result. Therefore, before processing each set of object parameters to be processed by the evaluation program, the set of object parameters to be processed can be processed in advance by a score prediction model to obtain a corresponding prediction result, and then the budget allocation value corresponding to each set of object parameters to be processed is determined according to the prediction result corresponding to each set of object parameters to be processed. Among them, the prediction result corresponding to each set of object parameters to be processed is proportional to its corresponding budget allocation value, that is, the larger the prediction result of the set of object parameters to be processed output by the score prediction model, the larger the budget allocated to it.

[0124] In a specific implementation, after generating the score prediction result corresponding to each parameter set of the object to be processed, the method further includes:

[0125] Invoking the object parameter generation algorithm in the search strategy module to process the object parameters to be adjusted in the target object parameter set to generate a set of object parameters to be evaluated corresponding to the target object;

[0126] The to-be-evaluated object parameter set is used as a to-be-processed object parameter set, and the budget allocation value corresponding to the target object parameter set determined by the budget control module is returned and executed.

[0127] Further, it can be determined whether a stop processing condition is met based on the processing information of the evaluation program;

[0128] If so, a set of object parameters for adjusting parameters of the target object is determined according to the score prediction result.

[0129] Specifically, the processing information of the evaluation program includes but is not limited to the running time and the output prediction results.

[0130] After the evaluation program outputs the score prediction result of the target object parameter set, the object parameter generation algorithm in the search strategy module can be re-called to process the object parameters to be adjusted in each target object parameter set, generate an object parameter set to be evaluated corresponding to the target object, and then use the object parameter set to be evaluated as the target object parameter set, and return to execute the process of determining the budget allocation value corresponding to the target object parameter set through the budget control module.

[0131] After obtaining the score prediction results output by the evaluation program, it is possible to determine whether the conditions for stopping processing are met based on the processing information of the evaluation program, for example, whether the running time of the evaluation program is greater than or equal to a preset duration threshold, or whether the performance of the target object determined by the score prediction results output by the evaluation program is greater than or equal to a preset score threshold; if so, the processing process of the score prediction model can be stopped.

[0132] This solution defines the components (base classes) and component interaction processes in the parameter tuning algorithm by combing and disassembling the parameter tuning algorithm process, so that the parameter tuning algorithm can also be developed in a modular way. And through the definition of Filter and Banker, the algorithm framework supports auxiliary means such as sensitive parameter identification and multi-fidelity algorithm to improve the performance of the parameter tuning algorithm.

[0133] The schematic diagram of the data processing process provided in the embodiments of this specification is as follows Figure 2b As shown in Figure 2, the overall parameter tuning process can be divided into two stages.

[0134] In the first stage, parameter value sampling and parameter sensitivity analysis are performed. First, the search space is initialized according to the Knobs configuration file, that is, the search space is generated. Then, the sampling method of the sampling module is called to sample the parameter values ​​of the object parameters of the target object, and multiple sampling points of each object parameter are generated. Based on the parameter values ​​corresponding to different sampling points of each object parameter, each object parameter is combined to generate multiple object parameter sets. Each object parameter set executes the evaluation procedure in turn to obtain an evaluation score (benchmark score). Then, according to the parameter value sampling points and benchmark scores of the object parameters in each object parameter set, the grant function of the parameter screening module is called to sort each object parameter set, and the object parameters of the target object are weighted according to the sorting results.

[0135] Through the first stage, we can obtain historical data consisting of multiple object parameter sets and their benchmark scores, as well as the ranking and weight of parameter sensitivity.

[0136] In the second stage, firstly, the historical data in the first stage (the object parameter set generated by combining each object parameter based on the parameter values ​​corresponding to different sampling points of each object parameter, and the evaluation score corresponding to each object parameter set) is used to train the score prediction model through the Fit function of the proxy module, and then the Generate and Evaluate functions of the search strategy module are used in turn, and based on the weights corresponding to the object parameters of the target object, the parameter values ​​of the object parameters to be adjusted are generated and screened, that is, candidate points of the object parameters to be adjusted are generated and screened, and different candidate points correspond to different parameter values, so as to combine each object parameter to be adjusted based on the parameter values ​​of the generated object parameters to be adjusted to generate a set of object parameters to be processed, thereby reducing the parameter dimension;

[0137] In addition, after the search strategy module generates the set of parameters of the objects to be processed, it can also filter the target object parameter set from the set of parameters of the objects to be processed. In this process, the score prediction model can be used to predict the benchmark score of the set of parameters of the objects to be processed, and the target object parameter set can be filtered according to the prediction results, so as to guide the screening process of the target object parameter set through the prediction results of the agent module.

[0138] After filtering and obtaining the target object parameter set, the Allocate function of the budget management module can be used to allocate a budget for each target object parameter set. The budget determines how much resources each target object parameter set can use to run the benchmark. Then, based on the budget allocation result, the evaluation program is used to predict the benchmark score of the target object parameter set. In this way, a cycle in stage 2 is completed. At the beginning of each cycle, the score prediction model can be adjusted with new data to ensure the accuracy of the output results of the score prediction model.

[0139] After one cycle is completed, the Generate and Evaluate functions of the search strategy module can be reused to generate and filter the parameter values ​​of the object parameters to be adjusted, that is, to generate new candidate points, and then combine the parameters of the object parameters to be adjusted based on the generated parameter values ​​of the object parameters to be adjusted to regenerate a new set of object parameters to be processed, and process it with the score prediction model to generate corresponding prediction scores, and then use the budget management module to allocate budgets to each set of object parameters to be processed according to the corresponding prediction scores, and continue to execute the above process in a cycle.

[0140] Until the end condition is triggered, the Best function of the search strategy module is used to determine a better object parameter set, so as to use the object parameter set to adjust the parameters of the target object.

[0141] This solution breaks down the parameter tuning algorithm process into multiple base classes and limits the interaction logic between the multiple base classes. Different base classes can have different implementations, and different implementations of different base classes can be arbitrarily combined to provide algorithm intelligent tuning services, which facilitates code iteration and maintenance, reduces the difficulty of secondary development, and is also conducive to increasing the participation of the open source community.

[0142] An embodiment of the present specification obtains parameter configuration information of a target object, samples parameter values ​​of the object parameters based on a parameter value range corresponding to the object parameters in the parameter configuration information, generates multiple object parameter sets according to the sampling results, wherein each object parameter set includes the object parameters, performs regression prediction on the target object based on the multiple object parameter sets, determines a target object parameter set from the multiple object parameter sets according to the prediction results, and performs parameter adjustment on the target object based on the target object parameter set.

[0143] The embodiments of this specification disassemble the parameter adjustment process of the target object so that the parameter adjustment algorithm development meets the principles of high cohesion and low coupling, which is beneficial to improving the maintainability and scalability of the parameter adjustment algorithm, and is also beneficial to reducing the difficulty of secondary development; in addition, by sampling the parameter values ​​of each object parameter and generating multiple object parameter sets based on the sampling results, and then selecting a target object parameter set from the multiple object parameter sets to adjust the parameters of the target object based on the target object parameter set, it is beneficial to improve the performance of the parameter adjustment algorithm, thereby helping to improve the accuracy of the parameter adjustment results of the target object.

[0144] The following combination Figure 3 , taking the application of the data processing method provided in this specification in a practical scenario as an example, the data processing method is further explained. Figure 3 A processing flow chart of a data processing method provided by an embodiment of the present specification is shown, which specifically includes the following steps.

[0145] Step 302: Acquire parameter configuration information of the target object, and initialize a search space based on the parameter configuration information of the target object, wherein the search space includes a parameter value range of an object parameter of the target object.

[0146] Step 304: sampling from the search space through a sampling module to obtain parameter value sampling results corresponding to the object parameters.

[0147] Specifically, the sampling module samples the parameter value of the object parameter based on the parameter value range corresponding to the object parameter in the search space.

[0148] Step 306: Generate multiple object parameter sets according to the sampling results, wherein each object parameter set includes a parameter value of an object parameter, or a combination of parameter values ​​corresponding to at least two object parameters.

[0149] Step 308: Perform regression prediction on the target object based on the multiple object parameter sets to generate corresponding prediction results.

[0150] Step 310: Use the multiple object parameter sets as training samples and the prediction results as labels to train the score prediction model to obtain the score prediction model.

[0151] Step 312: calling a screening function in a screening module, sorting the plurality of object parameter sets according to the prediction results, and setting weights for the object parameters according to the sorting results.

[0152] Step 314: Determine the object parameters to be adjusted from the object parameters according to the weight setting result.

[0153] Step 316: call the object parameter generation algorithm in the search strategy module to process the object parameters to be adjusted, and generate a set of object parameters to be processed corresponding to the target object.

[0154] Step 318: Determine the budget allocation value corresponding to each set of parameters of the object to be processed through the budget control module.

[0155] Step 320: Allocate a budget to each of the parameter sets of the object to be processed by using the budget allocation function in the budget control module.

[0156] Step 322: Based on the budget allocation result, an evaluation program is called to process each parameter set of the object to be processed, and a score prediction result corresponding to each parameter set of the object to be processed is generated.

[0157] Step 324 , calling the object parameter generation algorithm in the search strategy module to process the object parameters to be adjusted in each of the object parameter sets to be processed, and generating a set of object parameters to be evaluated corresponding to the target object.

[0158] The to-be-evaluated object parameter set is used as the to-be-processed object parameter set, and the process returns to step 318 .

[0159] Step 326, when it is determined according to the processing information of the evaluation program that the stop processing condition is met, the target object parameter set of the target object is determined according to the score prediction result, and the parameters of the target object are adjusted based on the parameter values ​​of each object parameter contained in the target object parameter set.

[0160] The embodiments of this specification disassemble the parameter adjustment process of the target object in a modular manner, so that the development of the parameter adjustment algorithm meets the principle of high cohesion and low coupling, which is beneficial to improving the maintainability and scalability of the parameter adjustment algorithm, and is also beneficial to reducing the difficulty of secondary development; in addition, by sampling the parameter values ​​of each object parameter and generating multiple object parameter sets based on the sampling results, and then selecting a target object parameter set from the multiple object parameter sets, so as to adjust the parameters of the target object based on the target object parameter set, it is beneficial to improve the performance of the parameter adjustment algorithm, thereby helping to improve the accuracy of the parameter adjustment results of the target object.

[0161] Corresponding to the above method embodiment, this specification also provides a data processing device embodiment, Figure 4 FIG. 1 is a schematic diagram showing the structure of a data processing device provided by an embodiment of the present specification. Figure 4 As shown, the device comprises:

[0162] The acquisition module 402 is configured to acquire parameter configuration information of the target object;

[0163] A sampling module 404 is configured to sample the parameter value of the object parameter based on the parameter value range corresponding to the object parameter in the parameter configuration information;

[0164] A generating module 406 is configured to generate a plurality of object parameter sets according to the sampling result, wherein each object parameter set includes a parameter value of an object parameter, or includes a combination of parameter values ​​corresponding to at least two object parameters respectively;

[0165] The adjustment module 408 is configured to perform regression prediction on the target object based on the multiple object parameter sets, determine a target object parameter set from the multiple object parameter sets according to the prediction result, and adjust parameters of the target object based on the target object parameter set.

[0166] Optionally, the adjustment module 408 is further configured to:

[0167] Using the multiple object parameter sets as training samples and the prediction results as labels, training the score prediction model to be trained to obtain the score prediction model;

[0168] A target object parameter set of the target object is determined according to the multiple object parameter sets and the score prediction model.

[0169] Optionally, the adjustment module 408 is further configured to:

[0170] Generate a to-be-processed object parameter set corresponding to the target object according to a preset algorithm and the plurality of object parameter sets;

[0171] Inputting the parameter set of the object to be processed into the score prediction model for processing, and generating a score prediction result corresponding to each parameter set of the object to be processed;

[0172] The target object parameter set is determined from the to-be-processed object parameter set according to the score prediction result.

[0173] Optionally, the adjustment module 408 is further configured to:

[0174] Sorting the plurality of object parameter sets according to the prediction results, and setting weights for the object parameters according to the sorting results;

[0175] Determining the object parameters to be adjusted from the object parameters according to the weight setting result;

[0176] According to the preset algorithm and the parameters of the object to be adjusted, a set of parameters of the object to be processed corresponding to the target object is generated.

[0177] Optionally, the data processing device further includes:

[0178] The initialization module is configured to initialize a search space based on parameter configuration information of a target object, wherein the search space includes a parameter value range of an object parameter of the target object.

[0179] Optionally, the sampling module 404 is further configured to:

[0180] The sampling module samples from the search space to obtain a parameter value sampling result corresponding to the object parameter.

[0181] Optionally, the adjustment module 408 is further configured to:

[0182] Calling a screening function in a screening module, sorting the plurality of object parameter sets according to the prediction results, and setting weights for the object parameters according to the sorting results;

[0183] Determining the object parameters to be adjusted from the object parameters according to the weight setting result;

[0184] The object parameter generation algorithm in the search strategy module is called to process the object parameters to be adjusted, and generate a set of object parameters to be processed corresponding to the target object.

[0185] Optionally, the data processing device further includes a first processing module configured to:

[0186] Determine the budget allocation value corresponding to the target object parameter set by a budget control module;

[0187] Allocating a budget to the target object parameter set by using a budget allocation function in the budget control module;

[0188] Based on the budget allocation result, an evaluation program is called to process the target object parameter set to generate a score prediction result corresponding to the target object parameter set.

[0189] Optionally, the data processing device further includes a second processing module configured to:

[0190] Invoking the object parameter generation algorithm in the search strategy module to process the object parameters to be adjusted in the target object parameter set to generate a set of object parameters to be evaluated corresponding to the target object;

[0191] The to-be-evaluated object parameter set is used as a to-be-processed object parameter set, and the budget allocation value corresponding to the target object parameter set is determined by the budget control module.

[0192] Optionally, the data processing device further includes a third processing module configured to:

[0193] determining whether a stop processing condition is met based on the processing information of the evaluation program;

[0194] If so, a set of object parameters for adjusting parameters of the target object is determined according to the score prediction result.

[0195] The above is a schematic scheme of a data processing device of this embodiment. It should be noted that the technical scheme of the data processing device and the technical scheme of the above data processing method belong to the same concept, and the details of the technical scheme of the data processing device that are not described in detail can be referred to the description of the technical scheme of the above data processing method.

[0196] Figure 5 The structure block diagram of a computing device 500 provided according to an embodiment of the present specification is shown. The components of the computing device 500 include but are not limited to a memory 510 and a processor 520. The processor 520 is connected to the memory 510 via a bus 530, and the database 550 is used to store data.

[0197] The computing device 500 also includes an access device 540 that enables the computing device 500 to communicate via one or more networks 560. Examples of these networks include a public switched telephone network (PSTN), a local area network (LAN), a wide area network (WAN), a personal area network (PAN), or a combination of communication networks such as the Internet. The access device 540 may include one or more of any type of network interface (e.g., a network interface card (NIC)) whether wired or wireless, such as an IEEE 802.11 wireless local area network (WLAN) wireless interface, a World Wide Interoperability for Microwave Access (Wi-MAX) interface, an Ethernet interface, a universal serial bus (USB) interface, a cellular network interface, a Bluetooth interface, a near field communication (NFC) interface, and the like.

[0198] In one embodiment of the present specification, the above components of the computing device 500 and Figure 5 Other components not shown in the figure may also be connected to each other, for example, via a bus. It should be understood that Figure 5 The computing device structure block diagram shown is only for the purpose of illustration, and is not intended to limit the scope of this specification. Those skilled in the art can add or replace other components as needed.

[0199] The computing device 500 may be any type of stationary or mobile computing device, including a mobile computer or mobile computing device (e.g., a tablet computer, a personal digital assistant, a laptop computer, a notebook computer, a netbook, etc.), a mobile phone (e.g., a smart phone), a wearable computing device (e.g., a smart watch, smart glasses, etc.), or other types of mobile devices, or a stationary computing device such as a desktop computer or PC. The computing device 500 may also be a mobile or stationary server.

[0200] The processor 520 is used to execute the following computer executable instructions, which implement the steps of the above data processing method when executed by the processor.

[0201] The above is a schematic scheme of a computing device of this embodiment. It should be noted that the technical scheme of the computing device and the technical scheme of the above data processing method belong to the same concept, and the details not described in detail in the technical scheme of the computing device can be referred to the description of the technical scheme of the above data processing method.

[0202] An embodiment of the present specification further provides a computer-readable storage medium storing computer-executable instructions, which can implement the steps of the above-mentioned data processing method when executed by a processor.

[0203] The above is a schematic scheme of a computer-readable storage medium of this embodiment. It should be noted that the technical scheme of the storage medium and the technical scheme of the above data processing method belong to the same concept, and the details not described in detail in the technical scheme of the storage medium can be referred to the description of the technical scheme of the above data processing method.

[0204] An embodiment of the present specification further provides a computer program, wherein when the computer program is executed in a computer, the computer is caused to execute the steps of the above-mentioned data processing method.

[0205] The above is a schematic scheme of a computer program of this embodiment. It should be noted that the technical scheme of the computer program and the technical scheme of the above data processing method belong to the same concept, and the details not described in detail in the technical scheme of the computer program can be referred to the description of the technical scheme of the above data processing method.

[0206] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0207] The computer instructions include computer program codes, which may be in source code form, object code form, executable files or some intermediate forms, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium may be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0208] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the embodiments of this specification are not limited by the order of the actions described, because according to the embodiments of this specification, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the embodiments of this specification.

[0209] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0210] The preferred embodiments of this specification disclosed above are only used to help explain this specification. The optional embodiments do not describe all the details in detail, nor do they limit the invention to only the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of the embodiments of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the embodiments of this specification, so that technicians in the relevant technical field can well understand and use this specification. This specification is only limited by the claims and their full scope and equivalents.

Claims

1. A data processing method, comprising: Get the parameter configuration information of the target object; Based on the parameter value range corresponding to the object parameter in the parameter configuration information, sampling the parameter value of the object parameter; Generate multiple object parameter sets according to the sampling results, wherein each object parameter set includes a parameter value of an object parameter, or includes a combination of parameter values ​​corresponding to at least two object parameters respectively; Performing regression prediction on the target object based on the multiple object parameter sets, determining a target object parameter set from the multiple object parameter sets according to the prediction result, and adjusting parameters of the target object based on the target object parameter set; Among them, determining the target object parameter set from the multiple object parameter sets according to the prediction results includes: predicting the score prediction results corresponding to each of the object parameter sets respectively through a score prediction model, and determining the target object parameter set according to the score prediction results, wherein the score prediction model is trained based on the multiple object parameter sets and the prediction results.

2. The data processing method according to claim 1, wherein the step of predicting the score prediction results corresponding to each of the object parameter sets by using the score prediction model, and determining the target object parameter set according to the score prediction results comprises: Using the multiple object parameter sets as training samples and the prediction results as labels, training the score prediction model to be trained to obtain the score prediction model; A target object parameter set of the target object is determined according to the multiple object parameter sets and the score prediction model.

3. The data processing method according to claim 2, wherein determining the target object parameter set according to the plurality of object parameter sets and the score prediction model comprises: Generate a to-be-processed object parameter set corresponding to the target object according to a preset algorithm and the plurality of object parameter sets; Inputting the parameter set of the object to be processed into the score prediction model for processing, and generating a score prediction result corresponding to each parameter set of the object to be processed; The target object parameter set is determined from the to-be-processed object parameter set according to the score prediction result.

4. The data processing method according to claim 3, wherein generating a set of object parameters to be processed corresponding to the target object according to a preset algorithm and the plurality of object parameter sets comprises: Sorting the plurality of object parameter sets according to the prediction results, and setting weights for the object parameters according to the sorting results; Determining the object parameters to be adjusted from the object parameters according to the weight setting result; According to the preset algorithm and the parameters of the object to be adjusted, a set of parameters of the object to be processed corresponding to the target object is generated.

5. The data processing method according to claim 1, after obtaining the parameter configuration information of the target object, further comprising: A search space is initialized based on parameter configuration information of a target object, wherein the search space includes a parameter value range of an object parameter of the target object.

6. The data processing method according to claim 5, wherein sampling the parameter value of the object parameter based on the parameter value range corresponding to the object parameter in the parameter configuration information comprises: The sampling module samples from the search space to obtain a parameter value sampling result corresponding to the object parameter.

7. The data processing method according to claim 3, wherein generating a set of object parameters to be processed corresponding to the target object according to a preset algorithm and the plurality of object parameter sets comprises: Calling a screening function in a screening module, sorting the plurality of object parameter sets according to the prediction results, and setting weights for the object parameters according to the sorting results; Determining the object parameters to be adjusted from the object parameters according to the weight setting result; The object parameter generation algorithm in the search strategy module is called to process the object parameters to be adjusted, and generate a set of object parameters to be processed corresponding to the target object.

8. The data processing method according to claim 3, after determining the target object parameter set from the to-be-processed object parameter set according to the score prediction result, further comprising: Determine the budget allocation value corresponding to the target object parameter set by a budget control module; Allocating a budget to the target object parameter set by using a budget allocation function in the budget control module; Based on the budget allocation result, an evaluation program is called to process the target object parameter set to generate a score prediction result corresponding to the target object parameter set.

9. The data processing method according to claim 8, after generating the score prediction result corresponding to the target object parameter set, further comprising: Invoking the object parameter generation algorithm in the search strategy module to process the object parameters to be adjusted in the target object parameter set to generate a set of object parameters to be evaluated corresponding to the target object; The to-be-evaluated object parameter set is used as the to-be-processed object parameter set, and the step of determining the budget allocation value corresponding to the target object parameter set by the budget control module is returned to be executed.

10. The data processing method according to claim 9, further comprising: determining whether a stop processing condition is met based on the processing information of the evaluation program; If so, a set of object parameters for adjusting parameters of the target object is determined according to the score prediction result.

11. A data processing device, comprising: An acquisition module, configured to acquire parameter configuration information of a target object; a sampling module configured to sample a parameter value of the object parameter based on a parameter value range corresponding to the object parameter in the parameter configuration information; A generating module is configured to generate a plurality of object parameter sets according to the sampling results, wherein each object parameter set includes a parameter value of an object parameter, or includes a combination of parameter values ​​corresponding to at least two object parameters respectively; an adjustment module, configured to perform regression prediction on the target object based on the multiple object parameter sets, determine a target object parameter set from the multiple object parameter sets according to the prediction result, and perform parameter adjustment on the target object based on the target object parameter set; The adjustment module is further configured to predict the score prediction results corresponding to each of the object parameter sets through a score prediction model, and determine the target object parameter set according to the score prediction results, wherein the score prediction model is trained based on multiple object parameter sets and the prediction results.

12. A computing device comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the data processing method according to any one of claims 1 to 10 are implemented.

13. A computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the data processing method according to any one of claims 1 to 10.

14. A computer program product, characterized in that The method comprises computer instructions which, when executed by a processor, implement the steps of the data processing method according to any one of claims 1 to 10.

Images