Method, device and electronic device for generating business data processing model

Abstract

The present disclosure provides a method, device and electronic device for generating a business data processing model, which relates to the field of artificial intelligence technology, specifically to technical fields such as deep learning and computer vision, and includes: determining model evaluation information of an initial business data processing model, determining a target operator based on the model evaluation information, determining target description information of the target operator based on the model evaluation information, and generating a target business data processing model based on the target description information and the target operator. Thus, the target operator and the target description information can be respectively determined based on the model evaluation information, thereby decoupling the determination process of the target operator and the target description information, effectively improving the flexibility of the target business data processing model generation process, and improving the target business data processing model generation efficiency.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, specifically to technical fields such as deep learning and computer vision, and in particular to a method, device and electronic device for generating a business data processing model. Background Art

[0002] Artificial intelligence is a discipline that studies how to use computers to simulate certain human thought processes and intelligent behaviors (such as learning, reasoning, thinking, planning, etc.). It includes both hardware-level and software-level technologies. Artificial intelligence hardware technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, and big data processing; artificial intelligence software technologies mainly include computer vision technology, speech recognition technology, natural language processing technology, as well as machine learning / deep learning, big data processing technology, knowledge graph technology, and other major directions.

[0003] In the related art, when generating a business data processing model, the efficiency of determining the operators required for the model and the operator description information (operator description information such as the operator's hyperparameters) is low, resulting in an inflexible generation process of the target business data processing model, affecting the generation efficiency of the target business data processing model. Summary of the invention

[0004] The present disclosure provides a method, device, electronic device, storage medium and computer program product for generating a business data processing model.

[0005] According to a first aspect of the present disclosure, a method for generating a business data processing model is provided, comprising: determining model evaluation information of an initial business data processing model; determining a target operator based on the model evaluation information; determining target description information of the target operator based on the model evaluation information; and generating a target business data processing model based on the target description information and the target operator.

[0006] According to a second aspect of the present disclosure, a device for generating a business data processing model is provided, comprising: a first determination module for determining model evaluation information of an initial business data processing model; a second determination module for determining a target operator based on the model evaluation information; a third determination module for determining target description information of the target operator based on the model evaluation information; and a generation module for generating a target business data processing model based on the target description information and the target operator.

[0007] According to a third aspect of the present disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute a method for generating a business data processing model as in the first aspect of the present disclosure.

[0008] According to a fourth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, where the computer instructions are used to enable a computer to execute the method for generating a business data processing model as described in the first aspect of the present disclosure.

[0009] According to a fifth aspect of the present disclosure, a computer program product is provided, including a computer program, which implements the steps of the method for generating a business data processing model of the first aspect of the present disclosure when executed by a processor.

[0010] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0012] Figure 1 is a schematic diagram according to a first embodiment of the present disclosure;

[0013] Figure 2 is a schematic diagram according to a second embodiment of the present disclosure;

[0014] Figure 3 is a schematic diagram according to a third embodiment of the present disclosure;

[0015] Figure 4 is a schematic diagram according to a fourth embodiment of the present disclosure;

[0016] Figure 5 is a schematic diagram according to a fifth embodiment of the present disclosure;

[0017] Figure 6 is a schematic diagram according to a sixth embodiment of the present disclosure;

[0018] Figure 7 is a schematic diagram according to a seventh embodiment of the present disclosure;

[0019] Figure 8 is a block diagram of an exemplary device for generating a business data processing model in an embodiment of the present disclosure;

[0020] Fig. 9A schematic block diagram of an example electronic device that can be used to implement the method for generating a business data processing model disclosed in the present invention is shown. DETAILED DESCRIPTION

[0021] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0022] Figure 1 is a schematic diagram according to the first embodiment of the present disclosure.

[0023] It should be noted that the executor of the method for generating a business data processing model in this embodiment is a device for generating a business data processing model, which can be implemented by software and / or hardware, and can be configured in an electronic device, which may include but is not limited to a terminal, a server, etc.

[0024] The embodiments of the present disclosure relate to the field of artificial intelligence technology, and specifically to technical fields such as deep learning and computer vision.

[0025] Among them, Artificial Intelligence (AI) is a new technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence.

[0026] Deep learning is the process of learning the inherent laws and representation levels of sample data. The information obtained in the learning process is of great help in interpreting data such as text, images, and sounds. The ultimate goal of deep learning is to enable machines to have analytical learning capabilities like humans and to be able to recognize data such as text, images, and sounds.

[0027] Computer vision refers to the use of cameras and computers to replace the human eye to identify, track and measure targets, and further perform graphic processing so that the computer processing becomes an image that is more suitable for human eye observation or transmission to instrument detection.

[0028] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0029] like Figure 1 As shown, the method for generating the business data processing model includes:

[0030] S101: Determine model evaluation information of an initial business data processing model.

[0031] The model may be, for example, an artificial intelligence model, such as a neural network model, a machine learning model, etc.

[0032] The business data processing model may refer to a model used to process various business data.

[0033] Among them, the initial business data processing model can be a business data processing model obtained in the initial stage of training the artificial intelligence model, or, when multiple rounds of iterative training are performed on the artificial intelligence model, the business data processing model obtained in the previous round of iterative training can also be used as the initial business data processing model, or the initial business data processing model can also be a pre-configured initial business data processing model.

[0034] In the disclosed embodiment, the model performance of the initial business data processing model can be analyzed and evaluated to obtain model evaluation information, and then, the selection of operators and the setting of hyperparameters in the next round of iterative training can be guided based on the model evaluation information, without any restriction.

[0035] In the embodiments of the present disclosure, the initial business data processing model can be constructed by multiple operators, and the business data processing model constructed by the multiple operators can be trained. When the business data processing model is constructed based on multiple operators, the description information of each operator can also be determined (the description information is such as a hyperparameter), and then, each operator is configured based on the description information to obtain the business data processing model, and the subsequent business data processing model training process is carried out to obtain the initial business data processing model.

[0036] In the disclosed embodiment, during the process of training the initial business data processing model, the performance of the initial business data processing model can also be dynamically analyzed to obtain its model evaluation information for the selection of operators and setting of hyperparameters in the next round of iterations, and there is no restriction on this.

[0037] In the embodiment of the present disclosure, an operator search space can be pre-configured, and the operator search space can include multiple types of operators. A parameter search space can be set, and the parameter search space can include hyperparameters corresponding to various types of operators. Then, in each round of iteration, the target operator required for the current round of iteration can be selected from the operator search space, and the hyperparameters required for the current round of iteration (which can be used as target description information) can be selected from the parameter search space. There is no limitation on this.

[0038] In the disclosed embodiment, the number of the above-mentioned initial business data processing models may be one or more, and the business data that can be processed by the business data model may be, for example, financial risk control data, video data, image data, text data, audio data, etc., without any limitation.

[0039] Among them, model evaluation information may refer to the relevant information obtained by evaluating the attribute characteristics of one or more dimensions of the above-mentioned business data processing model, for example, it may be the work efficiency evaluation information, stability evaluation information, output accuracy evaluation information, etc. of the initial business data processing model, and there is no limitation on this.

[0040] In the embodiment of the present disclosure, when determining the model evaluation information of the initial business data processing model, it is possible to pre-establish a communication link between the execution subject of the embodiment of the present disclosure and the big data server, and then obtain the model evaluation information corresponding to the initial business data processing model from the big data server. Alternatively, it is possible to pre-acquire multiple reference business data, and then use the initial business data processing model to process the multiple reference business data, and then obtain the corresponding model evaluation information based on the analysis of the processing results. There is no limitation to this.

[0041] S102: Determine the target operator according to the model evaluation information.

[0042] Here, the operator, i.e., the machine learning algorithm, may refer to a process that runs on business data to create a business data processing model. The operator may be, for example, a K-nearest neighbor algorithm, a linear regression algorithm, etc., without limitation.

[0043] The target operator may refer to an operator selected based on the model evaluation information. Since the target operator may be selected with reference to the model evaluation information of the initial business data processing model, the obtained target operator may be applicable to the target business data processing model to be constructed in the current round.

[0044] In some embodiments, the target operator is determined based on the model evaluation information. An evaluation threshold may be pre-configured, and then a candidate operator is randomly selected. The candidate operator evaluation information corresponding to the candidate operator is determined based on the model evaluation information. The candidate operator evaluation information is analyzed and compared with the evaluation threshold. When the candidate operator evaluation information is greater than or equal to the evaluation threshold, the corresponding candidate operator is used as the target operator. There is no restriction on this.

[0045] In other embodiments, the target operator is determined based on the model evaluation information, and the operator characteristics of each optional operator are obtained. Then, the matching degree between the model evaluation information and each operator characteristic is analyzed, and the target operator is determined from multiple optional operators based on the matching degree. There is no limitation to this.

[0046] Of course, any other possible method may be used to determine the target operator based on the model evaluation information, such as engineering methods, combination of numbers and shapes, etc., and there is no limitation to this.

[0047] In the disclosed embodiment, since there may be multiple operators available for selection in the process of generating the target business data processing model, and the performance information corresponding to different operators may be different, when the target operator is determined based on the model evaluation information, the performance of the target operator can be guaranteed, thereby providing reliable reference information for the subsequent generation process of the target business data processing model.

[0048] S103: Determine target description information of the target operator according to the model evaluation information.

[0049] The description information may refer to the hyperparameters corresponding to the operator, that is, some parameters used to determine the model. Different hyperparameters result in different models (for example, convolutional neural network models with different numbers of layers are different models). Hyperparameters may be variables determined empirically or configured according to the application scenario, and there is no restriction on this.

[0050] For example, in deep learning, hyperparameters include: learning rate, number of iterations, number of layers, number of neurons in each layer, etc.

[0051] Among them, the target description information refers to the description information of the target operator determined based on the model evaluation information, which can be used to generate the target business data processing model. The target description information is, for example, the specified learning rate, number of iterations, number of model layers, etc.

[0052] In some embodiments, the target description information of the target operator is determined based on the model evaluation information. The initial description information of the target operator can be determined, and a corresponding initial vector is generated based on the initial description information. The initial vector is assigned a value and the value of the initial vector is changed. The target operator performance corresponding to the initial vectors with various values ​​is evaluated based on the model evaluation information, and corresponding evaluation values ​​are formed. The value of the initial vector corresponding to the maximum evaluation value is selected as the final optimized value of the initial vector, and then the initial description information is optimized based on the final optimized value to obtain the target description information.

[0053] In other embodiments, the target description information of the target operator is determined based on the model evaluation information, or any other possible method may be used to determine the target description information of the target operator based on the model evaluation information, such as determining the initial description information corresponding to the target operator and then optimizing it using a Bayesian optimization method to generate the target description information, without limitation.

[0054] It is understandable that the same operator can use different description information, and the model performance generated by different description information combined with the operator may be different. Determining the target description information of the target operator based on the model evaluation information can ensure the description effect of the obtained target description information and improve the reliability of the target description information in the process of generating the target business data processing model.

[0055] After determining the model evaluation information of the initial business data processing model, the disclosed embodiment can respectively determine the adapted target operator and target description information based on the model evaluation information, thereby achieving effective decoupling of the target operator and target description information determination process, while ensuring the contribution of the target operator and target description information to the model performance, effectively improving the flexibility of the target business data processing model generation process and improving the target business data processing model generation efficiency.

[0056] In the disclosed embodiment, there may be multiple target operators, and correspondingly, each target operator has corresponding target description information. In the process of determining the target operator according to the model evaluation information, multiple target operators and the target description information corresponding to each target operator may be determined simultaneously based on the model evaluation information, thereby greatly improving the search and selection efficiency of the target operators and the target description information.

[0057] S104: Generate a target business data processing model according to the target description information and the target operator.

[0058] The target business data processing model may refer to a business data processing model generated based on the target description information and the target operator.

[0059] That is to say, after determining the target description information and the target operator according to the model evaluation information, the embodiment of the present disclosure can perform description processing on the target operator based on the target description information, and generate a target business data processing model based on the described target operator.

[0060] In some embodiments, the number of target operators may be multiple. A target business data processing model may be generated based on the target description information and the target operators. Priority information of multiple target operators may be determined, and then the corresponding target operators may be described and processed using the target description information based on the priority information to obtain multiple described target operators, and the target business data processing model may be generated based on the multiple described target operators.

[0061] In other embodiments, a target business data processing model is generated based on the target description information and the target operator. One or more target operators may be randomly selected from multiple target operators as the target operators to be described, and the target operators to be described are described according to the target description information corresponding to the target operators to be described to obtain multiple described target operators, and a target business data processing model is generated based on the multiple described target operators. There is no limitation to this.

[0062] In this embodiment, by determining the model evaluation information of the initial business data processing model, determining the target operator according to the model evaluation information, determining the target description information of the target operator according to the model evaluation information, and generating the target business data processing model according to the target description information and the target operator, the target operator and the target description information can be determined respectively based on the model evaluation information, thereby decoupling the determination process of the target operator and the target description information, effectively improving the flexibility of the target business data processing model generation process, and improving the target business data processing model generation efficiency.

[0063] Figure 2 is a schematic diagram according to a second embodiment of the present disclosure.

[0064] like Figure 2 As shown, the method for generating the business data processing model includes:

[0065] S201: Determine model evaluation information of an initial business data processing model.

[0066] The description of S201 can be specifically referred to the above embodiment, which will not be repeated here.

[0067] S202: Determine multiple candidate operators.

[0068] The candidate operators refer to multiple operators that may be used as target operators.

[0069] In the disclosed embodiment, multiple candidate operators may be determined by predetermining the application scenario and characteristic information of the initial business data processing model, and determining multiple operators that may be applicable to the business data processing model as candidate operators based on the characteristic information, thereby providing a reliable analysis object for the subsequent determination of the target operator.

[0070] S203: Determine candidate operator evaluation information corresponding to each candidate operator according to the model evaluation information.

[0071] The candidate operator evaluation information refers to the evaluation information of the candidate operator determined based on the model evaluation information, and the candidate operator evaluation information can be used to evaluate the performance of the candidate operator to determine the performance difference between different candidate operators. The candidate operator evaluation information may include the number of iterations corresponding to the candidate operator, operator performance information, etc., which are not limited.

[0072] In some embodiments, candidate operator evaluation information corresponding to each candidate operator is determined based on the model evaluation information. The time resources and system memory resources consumed by each candidate operator during the execution process can be determined based on the model evaluation information, and the time resources and system memory resources are used as the corresponding candidate operator evaluation information.

[0073] In other embodiments, candidate operator evaluation information corresponding to each candidate operator is determined based on the model evaluation information. Any other possible information may also be determined as the candidate operator evaluation information based on the model evaluation information, such as the robustness of each candidate operator to abnormal inputs, the readability of each candidate operator itself, etc., without limitation.

[0074] In the disclosed embodiment, candidate operator evaluation information corresponding to each candidate operator is determined based on the model evaluation information. The obtained candidate operator evaluation information can effectively characterize the evaluation information of one or more dimensions of the corresponding candidate operator, and can provide a clear and accurate reference for the subsequent determination of the target operator from multiple candidate operators.

[0075] S204: Determine target operator evaluation information from multiple candidate operator evaluation information, and use the candidate operator corresponding to the target operator evaluation information as the target operator.

[0076] The target operator evaluation information may refer to one or more candidate operator evaluation information selected from multiple candidate operator evaluation information based on a preset condition. For example, it may be to determine the priority information of multiple candidate operator evaluation information, and then select one or more candidate operator evaluation information from the multiple candidate operator evaluation information as the target operator evaluation information according to the priority information.

[0077] In some embodiments, the candidate operator evaluation information may contain multiple evaluation information, and the multiple evaluation information corresponds to multiple dimensional features of the candidate operator. In this case, the target operator evaluation information is determined from the multiple candidate operator evaluation information. The multiple evaluation information in a single candidate operator evaluation information can be weighted to obtain comprehensive evaluation information of multiple dimensional features, and then the comprehensive evaluation information corresponding to the multiple candidate operator evaluation information is analyzed and compared, and the candidate operator evaluation information corresponding to the optimal comprehensive evaluation information is used as the target operator evaluation information.

[0078] In other embodiments, the target operator evaluation information is determined from multiple candidate operator evaluation information, and dimensional evaluation information corresponding to one or more dimensional features adapted to the application scenario is selected from the candidate operator evaluation information, and the dimensional evaluation information corresponding to the multiple candidate operators is analyzed and compared to determine the target operator evaluation information suitable for the application scenario.

[0079] In the disclosed embodiment, preferably, multiple candidate operators can be determined, and based on the model evaluation information, the candidate operator evaluation information corresponding to each candidate operator can be determined, and then the target operator evaluation information can be determined from the multiple candidate operator evaluation information, and the candidate operator corresponding to the target operator evaluation information can be used as the target operator. Thus, the obtained candidate operator evaluation information can effectively characterize the performance information of the corresponding candidate operator, and based on the candidate operator evaluation information, the candidate operator with better performance can be accurately and quickly determined from the multiple candidate operators as the target operator, thereby ensuring the reliability of the target operator determination process.

[0080] S205: Determine initial description information of the target operator.

[0081] The initial description information refers to the unprocessed description information of the target operator in the initial state. The initial description information may be, for example, the hyperparameter value initially configured for the target operator, or the preset value of the hyperparameter, without limitation.

[0082] In the embodiment of the present disclosure, the initial description information of the target operator can be determined by pre-establishing a communication link between the execution subject of the embodiment of the present disclosure and the big data server, and obtaining better description information of the target operator from the big data server as the initial description information, thereby, the initial description information can provide a reliable reference object for determining the target description information.

[0083] S206: Determine an information adjustment value of the initial description information according to the model evaluation information.

[0084] The information adjustment value refers to the adjustment information determined by the initial description information based on the model evaluation information. The information adjustment value can be used as a reference for subsequent adjustment of the initial description information.

[0085] The information adjustment value is, for example, a hyperparameter value to be initially configured for the target operator, or a value to be adjusted from a preset value of a hyperparameter, and there is no restriction on this.

[0086] In the disclosed embodiment, the information adjustment value of the initial description information is determined based on the model evaluation information. The performance trend of the initial description information under different values ​​can be determined based on the model evaluation information, and then the information adjustment value of the initial description information is determined based on the performance trend. This can provide an accurate execution basis for the subsequent adjustment process of the initial description information.

[0087] S207: Adjust the initial description information according to the information adjustment value to obtain the target description information.

[0088] That is to say, in the embodiments of the present disclosure, the initial description information can be flexibly adjusted based on the information adjustment value, thereby ensuring the adaptability of the obtained target description information to the target business data processing model construction process.

[0089] In the disclosed embodiment, preferably, the initial description information of the target operator can be determined, and based on the model evaluation information, the information adjustment value of the initial description information can be determined, and then the initial description information can be adjusted based on the information adjustment value to obtain the target description information. Since the information adjustment value is obtained based on the model evaluation information, and the model evaluation information can effectively characterize the performance information of the target operator, the obtained information adjustment value can provide a clear and reliable reference basis for the adjustment process of the initial description information, and can effectively improve the description effect of the obtained target description information on the target operator.

[0090] S208: Describing corresponding target operators according to the multiple target description information to obtain multiple described target operators.

[0091] The described target operator refers to the operator obtained through the description process of the target operator corresponding to the target description information.

[0092] In the disclosed embodiment, by describing the corresponding target operators according to multiple target description information, the synchronous description processing of multiple target operators can be achieved quickly and accurately, thereby effectively improving the work efficiency of building a target business data processing model.

[0093] S209: Construct a target business data processing model according to the multiple described target operators.

[0094] In the disclosed embodiment, by constructing a target business data processing model according to a plurality of described target operators, the performance of the target business data processing model can be effectively improved so that it can be suitable for different application environments.

[0095] In the embodiments of the present disclosure, preferably, corresponding target operators can be described in parallel according to multiple target description information to obtain multiple described target operators. A target business data processing model can be constructed based on the multiple described target operators. Parallel processing of the description processes corresponding to the multiple target operators can be achieved, which greatly reduces the time cost of the overall description process. Based on the multiple described target operators, the construction effect of the target business data processing model can be effectively improved, ensuring the working performance of the obtained target business data processing model in different application environments.

[0096] In this embodiment, by determining the model evaluation information of the initial business data processing model, multiple candidate operators are determined, and according to the model evaluation information, the candidate operator evaluation information corresponding to each candidate operator is determined, and the target operator evaluation information is determined from the multiple candidate operator evaluation information, and the candidate operator corresponding to the target operator evaluation information is used as the target operator. Thus, the obtained candidate operator evaluation information can effectively characterize the performance information of the corresponding candidate operator, and based on the candidate operator evaluation information, the candidate operator with better performance can be accurately and quickly determined from the multiple candidate operators as the target operator, thereby ensuring the reliability of the target operator determination process, and then determining the initial description information of the target operator, and according to the model evaluation information, determining the information adjustment value of the initial description information, and adjusting the initial description information according to the information adjustment value to obtain the target description information. The information adjustment value is obtained based on the model evaluation information, and the model evaluation information can effectively characterize the performance information of the target operator, so that the obtained information adjustment value can provide a clear and reliable reference basis for the adjustment process of the initial description information, and can effectively improve the description effect of the obtained target description information on the target operator. By describing the corresponding target operator according to multiple target description information to obtain multiple described target operators, and constructing a target business data processing model according to the multiple described target operators, it is possible to realize parallel processing of the corresponding description processes of multiple target operators, and greatly reduce the time cost of the overall description process. Based on the multiple described target operators, the construction effect of the target business data processing model can be effectively improved, and the working performance of the obtained target business data processing model in different application environments is guaranteed.

[0097] Figure 3 is a schematic diagram according to a third embodiment of the present disclosure.

[0098] like Figure 3 As shown, the method for generating the business data processing model includes:

[0099] S301: Determine model evaluation information of an initial business data processing model.

[0100] S302: Determine multiple candidate operators.

[0101] The description of S301 - S302 can be specifically referred to the above embodiment, which will not be repeated here.

[0102] S303: Determine the number of iterations for each candidate operator based on the model evaluation information.

[0103] It can be understood that before determining the target operator, the embodiments of the present disclosure may involve the above-mentioned multiple candidate operators in multiple rounds of model evaluation to improve the performance of each candidate operator. During each round of model evaluation, the candidate operators participating in the next round of model evaluation can be determined based on the number of model evaluation rounds and performance information of the multiple candidate operators.

[0104] The number of iterations refers to the number of times the candidate operator participates in the above-mentioned multiple rounds of model evaluation.

[0105] In the disclosed embodiment, candidate operators consume time cost and computing power cost when performing iterations, and the amount of time cost and computing power cost consumed may increase with the increase in the number of iterations. Therefore, when the number of iterations of each candidate operator is determined based on the model evaluation information, the number of iterations can effectively characterize the cost information of each candidate operator, thereby providing a reliable reference basis for the subsequent determination of the candidate operator evaluation information, and ensuring the comprehensiveness of the obtained candidate operator evaluation information.

[0106] S304: Determine operator performance information of each candidate operator according to the model evaluation information.

[0107] The operator performance information may refer to relevant information used to describe the performance of each candidate operator.

[0108] For example, a set of operator performance evaluation standards may be formulated in advance, and then a comparative analysis may be performed based on the model evaluation information corresponding to each candidate operator and the operator performance evaluation standard to determine the operator performance information of each candidate operator.

[0109] In the disclosed embodiment, by determining the operator performance information of each candidate operator based on the model evaluation information, a suitable reference basis can be provided for the subsequent determination of the candidate operator evaluation information, so that the candidate operator evaluation information can effectively characterize the performance information of each candidate operator.

[0110] S305: Use the number of iteration participation times and the operator performance information as candidate operator evaluation information corresponding to the corresponding candidate operator.

[0111] That is to say, after obtaining the number of iteration participations and the operator performance information, the disclosed embodiment can use the number of iteration participations and the operator performance information as the candidate operator evaluation information corresponding to the corresponding candidate operator, so as to achieve a comprehensive and accurate consideration of the corresponding candidate operator by the candidate operator evaluation information, and can effectively improve the characterization capability of the candidate operator evaluation information for the candidate operator-related information.

[0112] For example, there are 3 candidate operators (candidate operator 1, candidate operator 2, candidate operator 3). At the current time point, candidate operator 1 has been iterated 10 times (evaluation score is 0.1), candidate operator 2 has been iterated 5 times (evaluation score is 0.3), and candidate operator 3 has been iterated 2 times (evaluation score is 0.4). At this time, the performance of candidate operator 1 is better (evaluation score is 1), the performance of candidate operator 2 is second (evaluation score is 0.9), and the performance of candidate operator 3 is poor (evaluation score is 0.6). At the current time point, candidate operator 2 has the highest comprehensive evaluation score (score = 0.9 + 0.3 = 1.2), so candidate operator 2 can be selected as the target operator.

[0113] In the disclosed embodiments, preferably, the number of iterations of each candidate operator can be determined based on the model evaluation information, the operator performance information of each candidate operator can be determined based on the model evaluation information, and the number of iterations and the operator performance information can be used as the candidate operator evaluation information corresponding to the corresponding candidate operator. The obtained candidate operator evaluation information can be enriched to a great extent from the two dimensions of the number of iterations and the operator performance information, thereby achieving a comprehensive and accurate consideration of each candidate operator and improving the characterization capability of the candidate operator evaluation information for the candidate operator related information.

[0114] S306: Determine target operator evaluation information from multiple candidate operator evaluation information, and use the candidate operator corresponding to the target operator evaluation information as the target operator.

[0115] The description of S306 can be specifically referred to the above embodiment, which will not be repeated here.

[0116] S307: Determine multiple candidate description information of the target operator.

[0117] The candidate description information refers to multiple description information applicable to the target operator.

[0118] It is understandable that the target operator may correspond to multiple candidate description information, and the description effects of the multiple candidate description information on the target operator may be different.

[0119] In the disclosed embodiment, by determining a plurality of candidate description information of a target operator, a reliable analysis object can be provided for subsequent determination of the target description information.

[0120] S308: Determine the matching degree value between the model evaluation information and each candidate description information.

[0121] The matching degree value may refer to a numerical value obtained by matching each candidate description information with the model evaluation information, which is used to characterize the matching degree between each candidate description information and the model evaluation information.

[0122] In the disclosed embodiment, since the model evaluation information can effectively characterize the best description information of the target operator in the model evaluation process, when the matching degree value between the model evaluation information and each candidate description information is determined, the matching degree value provides an accurate judgment basis for the subsequent determination of the target description information.

[0123] S309: Select candidate description information corresponding to the maximum matching degree value from multiple matching degree values, and use the corresponding candidate description information as target description information.

[0124] That is to say, after determining the matching degree value between the model evaluation information and each candidate description information, the embodiment of the present disclosure can select the candidate description information corresponding to the largest matching degree value from multiple matching degree values, and use the corresponding candidate description information as the target description information to improve the applicability of the obtained target description information in the process of generating the target business data processing model.

[0125] In the disclosed embodiment, preferably, multiple candidate description information of the target operator can be determined, the matching degree value between the model evaluation information and each candidate description information can be determined, and then the candidate description information corresponding to the maximum matching degree value is selected from the multiple matching degree values, and the corresponding candidate description information is used as the target description information. Since the obtained matching degree value can effectively characterize the matching degree between the best description information obtained by the target operator during the model training process and each candidate description information, when the target description information is determined based on the maximum matching degree value, the adaptability of the obtained target description information in the target business data processing model generation process can be effectively improved, and the description effect of the obtained target description information can be maximized.

[0126] S310: Generate a target business data processing model according to the target description information and the target operator.

[0127] The description of S310 can be specifically referred to the above embodiment, which will not be repeated here.

[0128] In this embodiment, by determining the model evaluation information of the initial business data processing model, multiple candidate operators are determined, the number of iteration participations of each candidate operator is determined according to the model evaluation information, the operator performance information of each candidate operator is determined according to the model evaluation information, and the number of iteration participations and the operator performance information are used as the candidate operator evaluation information corresponding to the corresponding candidate operator. The obtained candidate operator evaluation information can be enriched to a great extent from the two dimensions of the number of iteration participations and the operator performance information, so as to achieve a comprehensive and accurate consideration of each candidate operator, and improve the characterization ability of the candidate operator evaluation information for the candidate operator related information, and then determine the target operator evaluation information from the multiple candidate operator evaluation information, and use the target operator evaluation information as the candidate operator evaluation information corresponding to the corresponding candidate operator. The candidate operator corresponding to the evaluation information is taken as the target operator, multiple candidate description information of the target operator is determined, the matching degree value between the model evaluation information and each candidate description information is determined, the candidate description information corresponding to the maximum matching degree value is selected from multiple matching degree values, and the corresponding candidate description information is taken as the target description information. Since the obtained matching degree value can effectively characterize the matching degree between the best description information obtained by the target operator in the model training process and each candidate description information, when the target description information is determined based on the maximum matching degree value, the adaptability of the obtained target description information in the target business data processing model generation process can be effectively improved, and the description effect of the obtained target description information can be maximized.

[0129] Figure 4 is a schematic diagram according to a fourth embodiment of the present disclosure.

[0130] like Figure 4 As shown, the generating device 40 of the business data processing model includes:

[0131] A first determination module 401 is used to determine model evaluation information of an initial business data processing model;

[0132] A second determination module 402 is used to determine a target operator according to the model evaluation information;

[0133] A third determination module 403 is used to determine target description information of the target operator according to the model evaluation information; and

[0134] The generation module 404 is used to generate a target business data processing model according to the target description information and the target operator.

[0135] In some embodiments of the present disclosure, Figure 5 As shown, Figure 5 is a schematic diagram according to the fifth embodiment of the present disclosure, the device 50 for generating a business data processing model comprises: a first determining module 501, a second determining module 502, a third determining module 503, and a generating module 504, wherein the second determining module 502 comprises:

[0136] A first determination submodule 5021 is used to determine multiple candidate operators;

[0137] A second determination submodule 5022 is used to determine candidate operator evaluation information corresponding to each candidate operator according to the model evaluation information;

[0138] The third determination submodule 5023 is used to determine the target operator evaluation information from multiple candidate operator evaluation information, and use the candidate operator corresponding to the target operator evaluation information as the target operator.

[0139] In some embodiments of the present disclosure, the third determining module 503 is specifically configured to:

[0140] Determine the initial description information of the target operator;

[0141] Determining information adjustment values ​​for the initial descriptive information based on the model evaluation information; and

[0142] The initial description information is adjusted according to the information adjustment value to obtain the target description information.

[0143] In some embodiments of the present disclosure, the third determining module 503 is further configured to:

[0144] Determine multiple candidate description information of the target operator;

[0145] Determining a matching degree value between the model evaluation information and each candidate description information; and

[0146] The candidate description information corresponding to the maximum matching degree value is selected from the multiple matching degree values, and the corresponding candidate description information is used as the target description information.

[0147] In some embodiments of the present disclosure, the second determining submodule 5022 is specifically configured to:

[0148] According to the model evaluation information, determine the number of iterations for each candidate operator;

[0149] Determining operator performance information of each candidate operator based on the model evaluation information; and

[0150] The number of iteration participations and operator performance information are used as candidate operator evaluation information corresponding to the corresponding candidate operator.

[0151] In some embodiments of the present disclosure, the number of target operators is multiple;

[0152] The generating module 504 is specifically used for:

[0153] Describing corresponding target operators according to multiple target description information to obtain multiple described target operators;

[0154] According to the multiple described target operators, a target business data processing model is constructed.

[0155] It is understandable that the present embodiment Figure 5 The business data processing model generating device 50 and the business data processing model generating device 40 in the above-mentioned embodiment, the first determination module 501 and the first determination module 401 in the above-mentioned embodiment, the second determination module 502 and the second determination module 402 in the above-mentioned embodiment, the third determination module 503 and the third determination module 403 in the above-mentioned embodiment, and the generating module 504 and the generating module 404 in the above-mentioned embodiment may have the same function and structure.

[0156] It should be noted that the aforementioned explanation of the method for generating a business data processing model is also applicable to the device for generating a business data processing model in this embodiment.

[0157] In this embodiment, by determining the model evaluation information of the initial business data processing model, determining the target operator according to the model evaluation information, determining the target description information of the target operator according to the model evaluation information, and generating the target business data processing model according to the target description information and the target operator, the target operator and the target description information can be determined respectively based on the model evaluation information, thereby decoupling the determination process of the target operator and the target description information, effectively improving the flexibility of the target business data processing model generation process, and improving the target business data processing model generation efficiency.

[0158] The method for generating a business data processing model in the embodiment of the present disclosure may be applied to the following device for generating a business data processing model. The framework structure of the device for generating a business data processing model may also be as follows:

[0159] Figure 6 is a schematic diagram according to a sixth embodiment of the present disclosure.

[0160] like Figure 6 As shown, the business data processing model generation device 60 may further include: an operator selection module 601 and an operator description module 602; wherein,

[0161] The operator selection module 601 is used to determine the model evaluation information of the initial business data processing model, determine the target operator according to the model evaluation information, and transmit the model evaluation information to the operator description module;

[0162] The operator description module 602 is used to determine the target description information of the target operator according to the model evaluation information, wherein the target operator and the target description information are used to generate a target business data processing model.

[0163] In some embodiments of the present disclosure, Figure 7 As shown, Figure 7 is a schematic diagram according to the seventh embodiment of the present disclosure. The device 70 for generating a business data processing model may further include: an operator selection module 701 and an operator description module 702, wherein the operator selection module 701 is specifically used to:

[0164] Determine multiple candidate operators;

[0165] Determining candidate operator evaluation information corresponding to each candidate operator according to the model evaluation information; and

[0166] The target operator evaluation information is determined from the plurality of candidate operator evaluation information, and the candidate operator corresponding to the target operator evaluation information is used as the target operator.

[0167] In the embodiment of the present disclosure, the operator selection module 701 can be used to schedule the model generation process, and thus, it can also be called a dispatch supervisor module (Dispatcher Supervisor). A process can be started in the dispatch supervisor module (DispatcherSupervisor) to receive and send messages based on the process, determine the target operator, and control a single dispatch module to perform corresponding operations according to the received instruction message.

[0168] In some embodiments of the present disclosure, the operator description module 702 is specifically used to:

[0169] Determine the initial description information of the target operator;

[0170] determining information adjustment values ​​for the initial descriptive information based on the model evaluation information; and

[0171] The initial description information is adjusted according to the information adjustment value to obtain the target description information.

[0172] In some embodiments of the present disclosure, the operator description module 702 is specifically used to:

[0173] Determine multiple candidate description information of the target operator;

[0174] Determining a matching degree value between the model evaluation information and each candidate description information; and

[0175] The candidate description information corresponding to the maximum matching degree value is selected from the multiple matching degree values, and the corresponding candidate description information is used as the target description information.

[0176] In the embodiment of the present disclosure, the operator description module 702 can be used to schedule and optimize the description information of a single target operator, and thus, it can also be called a single dispatching module (Single Dispatcher). A process can also be started in the single dispatching module (Single Dispatcher), and based on the process, the target operator can be initialized, the description information of the target operator can be tuned, and the evaluation indicators of the target operator can be collected.

[0177] It can be understood that the above-mentioned operator selection module 701 and operator description module 702 can be encapsulated in the tuner module (Tuner) at the same time to improve the integration level of the generation device of the business data processing model, and at the same time, the interaction efficiency between the operator selection module and the operator description module can be improved to ensure the working performance of the operator selection module and the operator description module.

[0178] In the process corresponding to the Tuner module, the scheduling supervisor submodule is the main thread, and the single scheduling submodule is the subthread. The scheduling supervisor submodule and the single scheduling submodule can communicate through a shared queue.

[0179] In some embodiments of the present disclosure, the operator selection module 701 is specifically used to:

[0180] According to the model evaluation information, determine the number of iterations for each candidate operator;

[0181] Determining operator performance information of each candidate operator based on the model evaluation information; and

[0182] The number of iteration participations and operator performance information are used as candidate operator evaluation information corresponding to the corresponding candidate operator.

[0183] In some embodiments of the present disclosure, the device further comprises:

[0184] The model generation module 703 is used to obtain a target operator identifier corresponding to the target operator, generate a model building message according to the target operator identifier, and transmit the model building message to the operator description module 702;

[0185] In the embodiment of the present disclosure, the model generation module 703 can be used to drive multiple processes of the business data processing model generation process, and thus, it can also be called a driver module (Driver). The driver module (Driver) can also start a process and receive messages, send messages, etc. based on the process.

[0186] It can be understood that in the generation device of the business data processing model, the above-mentioned Tuner module and Driver module communicate through interprocess communication (IPC), for example, it can be pipe communication, and the communication content can be composed of command and data. The transmission format is (command, data), command represents the type of command, and data represents the data object corresponding to the command operation.

[0187] The command types include but are not limited to:

[0188] #in

[0189] Initialize=b'IN', indicating initialization of the search space.

[0190] RequestTrialJobs=b'GE', indicating requesting an experiment.

[0191] ReportMetricData=b'ME', indicating reporting indicators.

[0192] UpdateSearchSpace=b'SS', indicating updating the search space.

[0193] ImportData=b'FD', which means importing data for breakpoint recovery.

[0194] TrialEnd=b'EN', indicating the end of a single experiment.

[0195] Terminate=b'TE', indicating the task is terminated.

[0196] #out

[0197] Initialized = b'ID', indicating that the search space has been initialized.

[0198] NewTrialJob=b'TR', indicating that the experimental configuration has been generated.

[0199] NoMoreTrialJobs=b'NO', indicating that there is no more configuration information in the search space.

[0200] KillTrialJob=b'KI', indicating termination of the experiment.

[0201] The data format may include but is not limited to:

[0202] #in

[0203] Initialize data: Used to initialize the search space of various machine learning algorithms.

[0204] The hyperparameter search space for multiple algorithms can be, for example,

[0205]

[0206] Among them, id represents different operators (types), and parameters represents the description information (type and value range) corresponding to the operator.

[0207] RequestTrialJobs data: The number of trials requested.

[0208] ReportMetricData data: Reported trial metrics.

[0209] UpdateSearchSpace data: updated search space.

[0210] ImportData data: Import the previous data list for experiment recovery.

[0211] TrialEnd data: reported trial status.

[0212] Terminate data: empty.

[0213] #out

[0214] Initialized data: empty.

[0215] NewTrialJob data:Trial configuration parameters.

[0216] NoMoreTrialJobs data: empty.

[0217] KillTrialJob data: experiment id.

[0218] Among them, the operator description module 702 is also used to receive a model construction message, and parse the model construction message to obtain a target operator identifier, and describe the target operator according to the target description information of the target operator to which the target operator identifier belongs, to obtain the described target operator, wherein the described target operator is used to build a target business data processing model.

[0219] In some embodiments of the present disclosure,

[0220] There are multiple operator description modules 702, and each operator description module 702 is used to determine the target description information of the corresponding target operator; wherein, multiple operator description modules 702 are used to describe the corresponding target operator to obtain multiple described target operators, and the multiple described target operators are jointly used to construct a target business data processing model.

[0221] In the disclosed embodiment, in the process of constructing the target business data processing model, the task processes of multiple operator description modules 702 are independent of each other, that is, the description information tuning process of multiple target operators can be implemented in parallel to improve the description information tuning efficiency of multiple target operators.

[0222] For example, Figure 8 is a block diagram of an exemplary device for generating a business data processing model in an embodiment of the present disclosure, such as Figure 8 As shown, the generating device of the exemplary business data processing model includes: a driver module (Driver) and a tuner module (Tuner), and the driver module and the tuner module can generate corresponding processes respectively. Among them, the tuner module can be divided into a dispatch supervisor submodule (Dispatcher Supervisor) and multiple single dispatch submodules (Single Dispatcher), and the number of the single dispatch submodule can be one or more, and there is no limitation on this.

[0223] Based on the generation device and communication protocol of the exemplary business data processing model, the experimental process of determining the target operator and the target description information corresponding to the target operator can be as follows:

[0224] ①The Driver module receives and forwards the message requesting initialization (Initialize); after the Dispatcher Supervisor of the Tuner module receives the Initialize request, it can initialize multiple Single Dispatchers (each Single Dispatcher corresponds to an operator). When all Single Dispatchers have completed initialization, the Dispatcher Supervisor can send a message to the Driver module indicating that initialization has been completed (Initialized); when the Driver module receives Initialized, the entire initialization task is completed.

[0225] ②The Driver module sends a message requesting an experiment (RequestTrialJobs); after receiving the RequestTrialJobs request, the Dispatcher Supervisor of the Tuner module can obtain the target operator that is more likely to produce better indicators in the next experiment for each experimental request based on the corresponding multi-operator selection strategy (flexibly supporting various strategies), thereby generating the next experimental configuration (target description information) in the corresponding Single Dispatcher, and enabling the Dispatcher Supervisor to send a new experiment request message (NewTrialJob) to the Driver module. The NewTrialJob information contains the target operator selected by the Dispatcher Supervisor and its target description information. When the Driver module receives the complete NewTrialJob information, it indicates that the requested experiment task is completed.

[0226] ③The Driver module receives and forwards the message requesting to report the metric data (ReportMetricData); when the Dispatcher Supervisor of the Tuner module receives the ReportMetricData request, based on the data information of ReportMetricData (including the operator type and operator evaluation information corresponding to the experiment), the corresponding Single Dispatcher can collect the operator evaluation information of the experiment (the hyperparameter tuning strategy will generate the corresponding operator description information for the next experiment request based on the collected operator evaluation information).

[0227] ④ After the stop condition is met (for example, the number of experiments reaches the preset number, or the time consumed exceeds the time budget, etc.), the Driver module receives and forwards the request to terminate the message (Terminate); after receiving the Terminate request, the DispatcherSupervisor of the Tuner module can send a Terminate signal to multiple SingleDispatchers in the Tuner module. When all Single Dispatcher threads stop running and the Dispatcher Supervisor thread also stops running, it indicates that the Tuner module process has stopped (that is, the target operator selection and the target description information optimization tasks corresponding to the target operator are completed).

[0228] It is understandable that the present embodiment Figure 7The business data processing model generating device 70 and the business data processing model generating device 60 in the above embodiment, the operator selection module 701 and the operator selection module 601 in the above embodiment, and the operator description module 702 and the operator description module 602 in the above embodiment may have the same function and structure.

[0229] It should be noted that the aforementioned explanation of the method for generating a business data processing model is also applicable to the device for generating a business data processing model in this embodiment.

[0230] In this embodiment, by determining the model evaluation information of the initial business data processing model, determining the target operator according to the model evaluation information, determining the target description information of the target operator according to the model evaluation information, and generating the target business data processing model according to the target description information and the target operator, the target operator and the target description information can be determined respectively based on the model evaluation information, thereby decoupling the determination process of the target operator and the target description information, effectively improving the flexibility of the target business data processing model generation process, and improving the target business data processing model generation efficiency.

[0231] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0232] Fig. 9 A schematic block diagram of an example electronic device that can be used to implement the method for generating a business data processing model of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0233] like Fig. 9 As shown, the device 900 includes a computing unit 901, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 902 or a computer program loaded from a storage unit 908 into a random access memory (RAM) 903. In the RAM 903, various programs and data required for the operation of the device 900 can also be stored. The computing unit 901, the ROM 902, and the RAM 903 are connected to each other via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0234] A number of components in the device 900 are connected to the I / O interface 905, including: an input unit 906, such as a keyboard, a mouse, etc.; an output unit 907, such as various types of displays, speakers, etc.; a storage unit 908, such as a disk, an optical disk, etc.; and a communication unit 909, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 909 allows the device 900 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0235] The computing unit 901 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 901 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 901 performs the various methods and processes described above, such as executing a method for generating a business data processing model. For example, in some embodiments, the method for generating a business data processing model may be implemented as a computer software program, which is tangibly included in a machine-readable medium, such as a storage unit 908. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 900 via the ROM 902 and / or the communication unit 909. When the computer program is loaded into the RAM 903 and executed by the computing unit 901, one or more steps of the method for generating the business data processing model described above may be executed. Alternatively, in other embodiments, the computing unit 901 may be configured to execute the method for generating a business data processing model in any other appropriate manner (for example, by means of firmware).

[0236] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0237] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0238] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, 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 foregoing.

[0239] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0240] The systems and techniques described herein can be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), the Internet, and a blockchain network.

[0241] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services ("Virtual Private Server", or "VPS" for short). The server may also be a server of a distributed system, or a server combined with a blockchain.

[0242] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0243] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A method for generating a business data processing model, comprising: Analyzing and evaluating the model performance of the initial business data processing model to determine model evaluation information of the initial business data processing model, wherein the business data processed by the initial business data processing model includes video data, image data, text data, and audio data; Determining a target operator according to the model evaluation information; Determining target description information of the target operator according to the model evaluation information; and Generate a target business data processing model according to the target description information and the target operator; Wherein, determining the target operator according to the model evaluation information includes: Determining application scenarios and characteristic information of the initial business data processing model; Determining multiple operators of the initial business data processing model as candidate operators based on the characteristic information; Determining candidate operator evaluation information corresponding to each of the candidate operators according to the model evaluation information; and The target operator evaluation information is determined from the plurality of candidate operator evaluation information, and the candidate operator corresponding to the target operator evaluation information is used as the target operator.

2. The method according to claim 1, wherein: The step of determining target description information of the target operator according to the model evaluation information includes: Determining initial description information of the target operator; determining, according to the model evaluation information, an information adjustment value of the initial description information; and The initial description information is adjusted according to the information adjustment value to obtain the target description information.

3. The method according to claim 1, wherein: The step of determining target description information of the target operator according to the model evaluation information includes: Determining multiple candidate description information of the target operator; Determining a matching degree value between the model evaluation information and each of the candidate description information; and The candidate description information corresponding to the maximum matching degree value is selected from the multiple matching degree values, and the corresponding candidate description information is used as the target description information.

4. The method according to claim 1, wherein: The step of determining candidate operator evaluation information corresponding to each of the candidate operators according to the model evaluation information includes: Determining the number of iterations of each candidate operator according to the model evaluation information; Determining operator performance information of each of the candidate operators according to the model evaluation information; and The number of iteration participations and the operator performance information are used as candidate operator evaluation information corresponding to the corresponding candidate operator.

5. The method according to any one of claims 1 to 4, wherein the number of the target operators is multiple; in, The generating a target business data processing model according to the target description information and the target operator includes: Describing corresponding target operators according to the plurality of target description information to obtain a plurality of described target operators; The target business data processing model is constructed according to the multiple described target operators.

6. A device for generating a business data processing model, comprising: A first determination module is used to analyze and evaluate the model performance of the initial business data processing model to determine the model evaluation information of the initial business data processing model, wherein the business data processed by the initial business data processing model includes video data, image data, text data, and audio data; A second determination module is used to determine a target operator according to the model evaluation information; A third determination module is used to determine the target description information of the target operator according to the model evaluation information; and A generating module, used for generating a target business data processing model according to the target description information and the target operator; Wherein, the second determining module includes: A first determination submodule is used to determine the application scenario and characteristic information of the initial business data processing model; and determine multiple operators of the initial business data processing model as candidate operators based on the characteristic information; a second determining submodule, configured to determine candidate operator evaluation information corresponding to each of the candidate operators according to the model evaluation information; and The third determination submodule is used to determine the target operator evaluation information from the multiple candidate operator evaluation information, and use the candidate operator corresponding to the target operator evaluation information as the target operator.

7. The device according to claim 6, wherein: The third determination module is specifically used to: Determining initial description information of the target operator; determining, according to the model evaluation information, an information adjustment value of the initial description information; and The initial description information is adjusted according to the information adjustment value to obtain the target description information.

8. The device according to claim 6, wherein: The third determining module is further used for: Determining multiple candidate description information of the target operator; Determining a matching degree value between the model evaluation information and each of the candidate description information; as well as The candidate description information corresponding to the maximum matching degree value is selected from the multiple matching degree values, and the corresponding candidate description information is used as the target description information.

9. The device according to claim 6, wherein: The second determining submodule is specifically used for: Determining the number of iterations of each candidate operator according to the model evaluation information; Determining operator performance information of each of the candidate operators according to the model evaluation information; as well as The number of iteration participations and the operator performance information are used as candidate operator evaluation information corresponding to the corresponding candidate operator.

10. The device according to any one of claims 6 to 9, wherein the number of the target operators is multiple; in, The generation module is specifically used for: Describing corresponding target operators according to the plurality of target description information to obtain a plurality of described target operators; The target business data processing model is constructed according to the multiple described target operators.

11. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 5.

12. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-5.

13. A computer program product comprising a computer program, which, when executed by a processor, implements the steps of the method according to any one of claims 1 to 5.

Images