Model training method and device based on cloud architecture, electronic equipment and storage medium

Through the cloud-based model training method, the characteristics and optimization algorithm of the target data set are determined, and the threshold and weight set are integrated, the problems of local optimal solution and computing efficiency of neural network models in prediction are solved, and the prediction accuracy and speed are improved.

CN119089969APending Publication Date: 2024-12-06ISOFTSTONE TECH SERVICE CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411127119.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-15
Publication Date
2024-12-06

Smart Images

  • Figure CN119089969A_ABST
    Figure CN119089969A_ABST
Patent Text Reader

Abstract

The invention discloses a model training method and device based on a cloud architecture, electronic equipment and a storage medium. The method comprises the steps of determining a data type of a target data set; determining a target optimization algorithm of the cloud architecture and a target model framework of the cloud architecture according to the data type of the target data set, and determining a reference threshold and a reference weight set according to the target optimization algorithm of the cloud architecture; determining a target feature set according to the target data set; determining a target threshold value and a target weight set according to the target model framework, the reference threshold value and the reference weight set of the cloud architecture, and fusing the target model framework, the target threshold value and the target weight set of the cloud architecture to obtain a reference model; and training a reference model according to the target feature set and determining a target model. According to the method, the target model is determined through the target model framework, the target threshold, the target weight set and the target feature set of the cloud architecture, the problem that an existing model is prone to generating a local optimal solution is solved, and the accuracy of model prediction is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a model training method, device, electronic device and storage medium based on cloud architecture. Background Art

[0002] At present, it has become a trend to apply neural network models to data prediction. However, different data types, different data fields and different prediction directions often correspond to one or more neural network models. Therefore, it is difficult to select the most suitable model for input data from multiple models. Using multiple models at the same time often takes a long time and requires high local configuration. Existing methods are difficult to balance time efficiency and calculation accuracy. Summary of the invention

[0003] The present invention provides a model training method, device, electronic device and storage medium based on cloud architecture to solve the problem that the model is prone to local optimal solutions, thereby affecting the prediction accuracy.

[0004] According to one aspect of the present invention, a model training method based on cloud architecture is provided, comprising:

[0005] Determine the data type of the target data set, the data type is used to characterize the field to which the target data set belongs and the prediction direction, the field to which the target data set belongs characterizes the type and data characteristics of the target data set, the prediction direction is the angle at which the target data set needs to be predicted, and the target data set is the data set that needs to be predicted;

[0006] Determine a target optimization algorithm for the cloud architecture and a target model framework for the cloud architecture according to the data type of the target data set, and determine a reference threshold and a reference weight set according to the target optimization algorithm for the cloud architecture, wherein the reference threshold and the reference weight set are a set consisting of thresholds and weight values ​​of the target model framework for the cloud architecture;

[0007] Determine a target feature set according to a target data set, wherein the target feature set is a set of common features selected from different feature sets of the target data set;

[0008] Determine a target threshold and a target weight set according to the target model framework, reference threshold and reference weight set of the cloud architecture, and fuse the target model framework, target threshold and target weight set of the cloud architecture to obtain a reference model;

[0009] The reference model is trained according to the target feature set and a target model is determined.

[0010] According to another aspect of the present invention, a model training device based on cloud architecture is provided, comprising:

[0011] A data type determination module is used to determine the data type of a target data set, wherein the data type is used to characterize the field to which the target data set belongs and the prediction direction, wherein the field to which the target data set belongs characterizes the type and data characteristics of the target data set, wherein the prediction direction is the angle at which the target data set needs to be predicted, and the target data set is the data set that needs to be predicted;

[0012] A framework determination module is used to determine a target optimization algorithm of a cloud architecture and a target model framework of the cloud architecture according to a data type of a target data set, and to determine a reference threshold and a reference weight set according to the target optimization algorithm of the cloud architecture, wherein the reference threshold and the reference weight set are a set consisting of a threshold and a weight value of the target model framework of the cloud architecture;

[0013] A feature set determination module, used to determine a target feature set according to a target data set, wherein the target feature set is a set of common features selected from different feature sets of the target data set;

[0014] A reference model determination module, used to determine a target threshold and a target weight set according to the target model framework, reference threshold and reference weight set of the cloud architecture, and to fuse the target model framework, target threshold and target weight set of the cloud architecture to obtain a reference model;

[0015] The target model determination module is used to train the reference model and determine the target model according to the target feature set.

[0016] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0017] at least one processor; and

[0018] a memory communicatively connected to the at least one processor; wherein,

[0019] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the cloud architecture-based model training method described in any embodiment of the present invention.

[0020] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the cloud architecture-based model training method described in any embodiment of the present invention when executed.

[0021] The technical solution of the embodiment of the present invention determines the data type of the target data set; determines the target optimization algorithm of the cloud architecture and the target model framework of the cloud architecture according to the data type of the target data set, and determines the reference threshold and reference weight set according to the target optimization algorithm of the cloud architecture. The reference threshold and reference weight set are determined by the optimization algorithm, which is helpful to find the threshold and weight value of the target model framework that best fits the cloud architecture; determines the target feature set according to the target data set, which is helpful to reduce the amount of data, grasp the key features, and reduce the difficulty of training, thereby improving the speed of calculation; determines the target threshold and target weight set according to the target model framework, reference threshold and reference weight set of the cloud architecture, and fuses the target model framework, target threshold and target weight set of the cloud architecture to obtain a reference model, which provides a selection basis for selecting the model that best fits the target data set, and the determination of the target threshold and target weight set is helpful to improve the convergence speed and find the global optimal solution; determines the target model according to the target feature set and the reference model. The method determines the target model through the target model framework, target threshold and target weight set, and target feature set of the cloud architecture, which can solve the problem that the existing neural network model determination method is prone to produce local optimal solutions, improves the accuracy of model prediction, and can also reduce the difficulty of model training and improve the calculation speed of the model.

[0022] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0024] Figure 1 A flowchart of a model training method based on cloud architecture provided in an embodiment of the present invention;

[0025] Figure 2 A flowchart of another cloud architecture-based model training method provided in an embodiment of the present invention;

[0026] Figure 3 A schematic diagram of the structure of a model training device based on a cloud architecture provided in an embodiment of the present invention;

[0027] Figure 4 A schematic diagram of the structure of an electronic device for implementing the cloud architecture-based model training method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0028] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0029] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0030] Figure 1 A flowchart of a cloud-based model training method provided in an embodiment of the present invention is provided. This embodiment is applicable to the case where the optimal neural network model is selected based on the cloud architecture according to the input data. The method can be executed by a cloud-based model training device. The cloud-based model training device can be implemented in the form of hardware and / or software. The cloud-based model training device can be configured in any electronic device with network communication function. Figure 1 As shown, the method includes:

[0031] S110: Determine the data type of the target data set.

[0032] Among them, the data type is used to characterize the field to which the target data set belongs and the prediction direction. The field characterizes the type and data characteristics of the target data set. The prediction direction is the angle at which the target data set needs to be predicted. The target data set is the data set that needs to be predicted.

[0033] Specifically, input the target data set in the data input interface, and select the field and prediction direction of the target data set.

[0034] For example, assuming that the input target data set is wheat growth data, the selected field is the agricultural field, the data type is text data, and the prediction direction is the growth trend or pest and disease status in the next XX days.

[0035] S120, determining a target optimization algorithm of the cloud architecture and a target model framework of the cloud architecture according to the data type of the target data set, and determining a reference threshold and a reference weight set according to the target optimization algorithm of the cloud architecture.

[0036] The reference threshold and reference weight set is a set consisting of the threshold and weight values ​​of the target model framework of the cloud architecture.

[0037] Specifically, historical case data is determined according to the data type of the target data set, and the target optimization algorithm of the corresponding cloud architecture is matched from the optimization algorithm library of the cloud architecture according to the historical case data. The target optimization algorithm of the cloud architecture is iteratively used to generate reference thresholds and reference weight sets, and the target model framework of the corresponding cloud architecture is matched from the model library of the cloud architecture.

[0038] Among them, the optimization algorithm library of the cloud architecture includes existing optimization algorithms, such as: target particle swarm optimization algorithm, locust optimization algorithm, butterfly optimization algorithm, elephant swarm optimization algorithm and chaotic seagull optimization algorithm, etc. The model library of the cloud architecture is composed of existing neural network models, such as: convolutional neural network model, recurrent neural network model, generative adversarial network model, graph neural network model, etc.

[0039] The above steps can improve the prediction accuracy by matching the appropriate model framework and optimization algorithm according to the field and prediction direction of the target dataset.

[0040] S130. Determine a target feature set according to the target data set.

[0041] The target feature set is a set of common features selected from different feature sets of the target data set.

[0042] Specifically, multiple feature selection algorithms are selected from the feature selection algorithm library of the cloud architecture according to the data type of the target data set. The feature selection algorithm performs feature selection on the target data set to obtain different feature sets. The common features in different feature sets are combined into a combination, and the combination is used as the target feature set.

[0043] Among them, the feature selection algorithm library of the cloud architecture is constructed by existing feature selection algorithms, such as filtering method, wrapping method, embedding method, etc.

[0044] Exemplarily, assuming that the target data set is text data, the corresponding selected feature selection algorithm is a method for processing text data.

[0045] In the above steps, the corresponding type of feature selection algorithm is selected according to the data characteristics of the target data set, which can improve the accuracy of feature extraction, and the selection of common features can eliminate the influence of unimportant features on the prediction results, reduce the difficulty of training, and improve the speed of training.

[0046] S140, determining a target threshold and a target weight set according to the target model framework, reference threshold and reference weight set of the cloud architecture, and fusing the target model framework, target threshold and target weight set of the cloud architecture to obtain a reference model.

[0047] Specifically, according to the target model framework of the cloud architecture, the corresponding threshold and weight set are matched from the reference threshold and reference weight set, and the matched threshold and weight set are used as the target threshold and target weight set. The target model framework of the cloud architecture, the target threshold and the target weight set are integrated to obtain the reference model.

[0048] The above steps match the target threshold and target weight set from the reference threshold and reference weight set, so that the threshold and weight set can match the target model framework of the cloud architecture to ensure the normal operation of the model.

[0049] S150: Train a reference model according to the target feature set and determine a target model.

[0050] Specifically, the target feature set is input into the reference model for calculation, the error of the predicted result is calculated, and the reference model with the smallest error is used as the target model.

[0051] Optionally, determining a target optimization algorithm of the cloud architecture and a target model framework of the cloud architecture according to the data type of the target data set includes steps A1-A3:

[0052] Step A1: Match the corresponding historical case data from the knowledge base cloud according to the data features of the target data set.

[0053] Among them, the historical case data includes: the model framework and optimization algorithm corresponding to the historical data set, and the historical data set and the target data set are data sets of the same type.

[0054] Specifically, according to the field to which the target data set belongs, the corresponding case data is matched from the knowledge base cloud according to the field, and the case data is filtered according to the prediction direction to obtain historical case data.

[0055] Step A2: Match the target model framework of the cloud architecture corresponding to the target data set from the model library of the cloud framework according to the historical case data.

[0056] Specifically, the target model framework of the corresponding cloud architecture is matched from the model library of the cloud framework according to the model framework adopted by the target data set in the historical case data.

[0057] For example, assuming that the target data set is wheat price increase data, and the prediction direction is the price increase in the next XX days, the historical case data of historical wheat price increase predictions are matched from the knowledge base cloud. Assuming that the models used in the historical case data include: linear regression model, long short-term memory network and BP neural network, etc., the model framework of linear regression model, long short-term memory network and BP neural network is matched from the model library of the cloud framework.

[0058] The above steps match the target model framework of the cloud architecture according to historical case data, solving the problem of difficult model selection.

[0059] Step A3: Match the target optimization algorithm of the cloud architecture from the optimization algorithm library of the cloud architecture according to the historical case data.

[0060] Specifically, the target model framework of the corresponding cloud architecture is matched from the optimization algorithm library of the cloud architecture according to the optimization algorithm adopted by the historical case data.

[0061] For example, assuming that the target data set is wheat price increase data, and the prediction direction is the price increase in the next XX days, the historical case data of wheat price increase predictions are matched from the knowledge base cloud. Assuming that the optimization algorithms used in the historical case data are genetic algorithms and ant colony algorithms, etc., the genetic algorithms and ant colony algorithms are matched from the optimization algorithm library of the cloud framework.

[0062] The above steps determine the model framework and optimization algorithm based on historical data, provide a basis for the selection of models and optimization algorithms, and can select a model framework that is more suitable for the target data set and an optimization algorithm that adapts to the model framework, thereby improving the accuracy of model prediction.

[0063] Optionally, determining a reference threshold and a reference weight set according to a target optimization algorithm of a cloud architecture includes steps B1-B3:

[0064] Step B1: Determine the network structure of the target model framework of the cloud architecture.

[0065] Specifically, the network structure is determined according to the composition of the target model framework of the cloud architecture and the number of inputs and outputs.

[0066] Exemplarily, assuming that the target model framework of the cloud architecture is a BP neural network, the corresponding network structure may be 1 input-3 hidden-1 output.

[0067] Step B2: Determine the objective function of the target optimization algorithm of the cloud architecture according to the network structure of the target model framework of the cloud architecture.

[0068] Specifically, the objective function of the target optimization algorithm of the cloud architecture is determined according to the number of inputs, the number of output layers, and the number of neurons of the target model framework of the cloud architecture.

[0069] Step B3: Determine reference thresholds and reference weight sets of target model frameworks of different cloud architectures according to the objective function.

[0070] Specifically, the initial threshold and initial weight value of the target model framework of the cloud architecture are generated by the target optimization algorithm of the cloud architecture, and the initial threshold and initial weight value are iterated, the threshold and weight value of the iteration result that infinitely approaches the target function are retained, and a reference threshold and reference weight set are generated.

[0071] For example, assume that there are N cloud architecture target optimization algorithms (O1, O2, O3...O N ), M cloud architecture target model frameworks (N1, N2, N3...N M ), the target optimization algorithms of N cloud architectures generate thresholds and weight values ​​of the target model frameworks of M cloud architectures respectively, and generate reference thresholds and reference weight sets (S1, S2, S3...S M×N ). Wherein, S1 represents the threshold and weight value of the target model framework N1 of the cloud architecture generated by the target optimization algorithm O1 of the cloud architecture.

[0072] The above steps determine the reference threshold and reference weight set through the target optimization algorithm of the cloud architecture according to the network structure of the target model framework of the cloud architecture, which can select the threshold and weight value of the target model framework of the appropriate cloud architecture, which is conducive to improving the model convergence speed and finding the global optimal solution.

[0073] Optionally, determining a target feature set according to a target data set includes steps C1-C3:

[0074] Step C1: Select multiple target feature selection algorithms from the feature selection algorithm library of the cloud architecture according to the data features of the target data set.

[0075] Specifically, according to the data features of the target data set, that is, whether the target data set is image data or text data, a target feature selection algorithm is determined from a feature selection algorithm library of the cloud architecture.

[0076] Step C2: extracting features from the target data set according to multiple target feature selection algorithms to obtain multiple reference feature sets.

[0077] Specifically, the target data set is input into a plurality of target feature selection algorithms, features of the target data set are extracted by the target feature selection algorithms, and the obtained feature data are used to generate a plurality of reference feature sets.

[0078] Step C3: Filter out common features from multiple reference feature sets, and form a target feature set with the common features.

[0079] Specifically, multiple reference feature sets are screened, common features therein are extracted, and the extracted common features are used to generate a target feature set.

[0080] In the above steps, the determination of the target feature set is conducive to screening out input features with high relevance, reducing the amount of data, grasping key features, reducing the difficulty of training and improving prediction accuracy.

[0081] Optionally, a target threshold and a target weight set are determined according to the target model framework, reference threshold and reference weight set of the cloud architecture, and the target model framework, target threshold and target weight set of the cloud architecture are integrated to obtain a reference model, including steps D1-D2:

[0082] Step D1: Match the corresponding target threshold and target weight set from the reference threshold and reference weight set according to the target model framework of the cloud architecture.

[0083] Specifically, according to the network structure of the target model framework of the cloud architecture, a target threshold and a target weight set corresponding to the network structure are matched from the reference threshold and the reference weight set.

[0084] For example, assume that there are N cloud architecture target optimization algorithms (O1, O2, O3...O N ), M cloud architecture target model frameworks (N1, N2, N3...N M ), reference threshold and reference weight set (S1, S2, S3...S M×N ), then the reference threshold and reference weight set S1-S N The target threshold and target weight set corresponding to the target model framework N1 of the cloud architecture, the target threshold and target weight set corresponding to the target model framework N2 of the cloud architecture, and so on.

[0085] The above steps of determining the target threshold and target weight set according to the target model framework of the cloud architecture help to select the threshold and weight set that are adapted to the target model framework of the cloud architecture from the reference threshold and reference weight set, avoiding the problem of incompatibility between the target model framework of the cloud architecture and the threshold and weight values ​​caused by random allocation according to the reference threshold and reference weight set when there are multiple target model frameworks of the cloud architecture.

[0086] Step D2: The target model framework, target threshold and target weight set of the cloud architecture are matched to obtain a reference model.

[0087] Specifically, the threshold values ​​and weight values ​​in the obtained target threshold value and target weight set are respectively matched to the target model framework of the cloud architecture to obtain a reference model.

[0088] For example, the target threshold and target weight set S1-S NBy corresponding to the target model framework N1 of the cloud architecture respectively, N reference models can be obtained.

[0089] In the above steps, the determination of the reference model helps to provide a basis for selecting the model that best suits the target data set.

[0090] Optionally, training a reference model according to the target feature set and determining a target model includes steps E1-E2:

[0091] Step E1: Use the target feature set as the input of the reference model, perform prediction through the reference model, and obtain a reference prediction result.

[0092] Specifically, the target feature set is input into the reference model, prediction is performed through the reference model, and the prediction result is used as the reference prediction result.

[0093] Step E2: determining a target model based on the reference prediction result and the target data set.

[0094] Specifically, the error between the reference prediction result and the target data set is calculated, and the target model is determined according to the size of the error.

[0095] Optionally, determining a target model according to a reference prediction result and a target data set includes steps F1-F2:

[0096] Step F1: determine an error index based on a reference prediction result and a target data set.

[0097] Specifically, the different error results between the reference prediction result and the target data set are calculated, and the obtained error results are used as error indicators.

[0098] Step F2: sort the error indicators and use the reference model corresponding to the minimum error indicator as the target model.

[0099] Specifically, the error indicators are sorted in descending order, and according to the sorting result, the reference model corresponding to the smallest error indicator is used as the target model.

[0100] For example, Figure 2As shown, first, the input data is input into the knowledge base cloud, and the corresponding historical case data is matched from the knowledge base cloud according to the data type of the input data. According to the historical case data, the corresponding target model framework of the cloud architecture is matched from the model library of the cloud architecture, that is, the neural network model library cloud. The target optimization algorithm of the cloud architecture is matched from the optimization algorithm library of the cloud architecture. According to the network structure of the target model framework of the cloud architecture and the data type of the input data, the target threshold and target weight set are generated by the target optimization algorithm of the cloud architecture. The threshold and weight values ​​in the target threshold and target weight set are matched with the target model framework of the cloud architecture. After the matching is completed, the reference model is obtained, and the reference model is trained according to the target feature set. The reference model with the smallest error is used as the target model, and the prediction result is output.

[0101] The technical solution of this embodiment determines the data type of the target data set; determines the target optimization algorithm of the cloud architecture and the target model framework of the cloud architecture according to the data type of the target data set, and determines the reference threshold and reference weight set according to the target optimization algorithm of the cloud architecture. The reference threshold and reference weight set are determined by the optimization algorithm, which is helpful to find the threshold and weight value of the target model framework that best suits the cloud architecture; determines the target feature set according to the target data set, which is helpful to reduce the amount of data, grasp the key features, reduce the difficulty of training, and thus improve the speed of calculation; determines the target threshold and target weight set according to the target model framework, reference threshold and reference weight set of the cloud architecture, and fuses the target model framework, target threshold and target weight set of the cloud architecture to obtain a reference model, which provides a selection basis for selecting the model that best suits the target data set, and the determination of the target threshold and target weight set is helpful to improve the convergence speed and find the global optimal solution; determines the target model according to the target feature set and the reference model. The method determines the target model through the target model framework, target threshold and target weight set, and target feature set of the cloud architecture, which can solve the problem that the existing neural network model determination method is prone to produce local optimal solutions, improves the accuracy of model prediction, and can also reduce the difficulty of model training and improve the calculation speed of the model.

[0102] Figure 3 The present invention provides a schematic diagram of a cloud-based model training device. This embodiment is applicable to the case where the optimal neural network model is selected based on input data based on the cloud architecture. The cloud-based model training device can be implemented in the form of hardware and / or software. The cloud-based model training device can be configured in any electronic device with network communication function. Figure 3 As shown, the device includes: a data type determination module 210, a framework determination module 220, a feature set determination module 230, a reference model determination module 240 and a target model determination module 250, wherein:

[0103] Data type determination module 210: used to determine the data type of the target data set. The data type is used to characterize the field to which the target data set belongs and the prediction direction. The field characterizes the type and data characteristics of the target data set. The prediction direction is the angle at which the target data set needs to be predicted. The target data set is the data set that needs to be predicted.

[0104] Framework determination module 220: used to determine the target optimization algorithm of the cloud architecture and the target model framework of the cloud architecture according to the data type of the target data set, and determine the reference threshold and reference weight set according to the target optimization algorithm of the cloud architecture, where the reference threshold and reference weight set are sets composed of thresholds and weight values ​​of the target model framework of the cloud architecture;

[0105] Feature set determination module 230: used to determine a target feature set according to a target data set, where the target feature set is a set of common features selected from different feature sets of the target data set;

[0106] Reference model determination module 240: used to determine a target threshold and a target weight set according to the target model framework, reference threshold and reference weight set of the cloud architecture, and fuse the target model framework, target threshold and target weight set of the cloud architecture to obtain a reference model;

[0107] The target model determination module 250 is used to train a reference model according to a target feature set and determine a target model.

[0108] Optionally, the framework determination module 220 includes:

[0109] Historical case data determination unit: used to match the corresponding historical case data from the knowledge base cloud according to the data characteristics of the target data set, the historical case data includes: the model framework and optimization algorithm corresponding to the historical data set, and the historical data set and the target data set are the same type of data sets;

[0110] A target model framework determination unit of a cloud architecture: used for matching a target model framework of a cloud architecture corresponding to a target data set from a model library of the cloud architecture according to historical case data;

[0111] A target optimization algorithm determination unit for a cloud architecture: used to match a target optimization algorithm for a cloud architecture from an optimization algorithm library for the cloud architecture according to historical case data.

[0112] Optionally, the framework determination module 220 includes:

[0113] Network structure determination unit: used to determine the network structure of the target model framework of the cloud architecture;

[0114] Weight determination unit: used to determine the objective function of the target optimization algorithm of the cloud architecture according to the network structure of the target model framework of the cloud architecture;

[0115] Reference weight set determination unit: used to determine reference thresholds and reference weight sets of target model frameworks of different cloud architectures according to the objective function.

[0116] Optionally, the feature set determination module 230 includes:

[0117] A selection algorithm determination unit: used for selecting a plurality of target feature selection algorithms from a feature selection algorithm library of a cloud architecture according to data features of a target data set;

[0118] A reference feature set determination unit is used to extract features from a target data set according to multiple target feature selection algorithms to obtain multiple reference feature sets;

[0119] Target feature set selection algorithm determination unit: used to screen out common features from multiple reference feature sets and form the common features into a target feature set.

[0120] Optionally, the reference model determination module 240 includes:

[0121] Target threshold and target weight set matching unit: used to match the corresponding target threshold and target weight set from the reference threshold and reference weight set according to the target model framework of the cloud architecture,

[0122] Reference model determination unit: used to match the target model framework of the cloud architecture with the target threshold and target weight set to obtain a reference model.

[0123] Optionally, the target model determination module 250 includes:

[0124] Reference prediction result determination unit: used to take the target feature set as the input of the reference model, perform prediction through the reference model, and obtain a reference prediction result;

[0125] Target model determination unit: used to determine the target model according to the reference prediction result and the target data set.

[0126] Optionally, the target model determination unit includes:

[0127] Error index determination subunit: used to determine the error index according to the reference prediction result and the target data set;

[0128] Target model determination subunit: used to sort the error indicators and take the reference model corresponding to the minimum error indicator as the target model.

[0129] The cloud-based model training device provided in the embodiments of the present invention can execute the cloud-based model training method provided in any of the above embodiments of the present invention, and has the corresponding functions and beneficial effects of executing the cloud-based model training method. For detailed processes, please refer to the relevant operations of the cloud-based model training method in the above embodiments.

[0130] Figure 4 A schematic diagram of the structure of an electronic device for implementing a cloud-based model training method according to an embodiment of the present invention. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.), 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 invention described and / or required herein.

[0131] like Figure 4 As shown, the electronic device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0132] A number of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0133] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as a model training method based on a cloud architecture.

[0134] In some embodiments, the cloud-based model training method may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the cloud-based model training method described above may be performed. Alternatively, in other embodiments, the processor 11 may be configured to execute the cloud-based model training method in any other appropriate manner (e.g., by means of firmware).

[0135] 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.

[0136] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program 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.

[0137] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. 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.

[0138] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device 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 a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0139] The systems and techniques described herein may 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 may 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), a blockchain network, and the Internet.

[0140] A computing 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 client and server relationship 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.

[0141] 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 described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0142] The above specific implementations do not constitute a limitation on the protection scope of the present invention. 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 invention should be included in the protection scope of the present invention.

Claims

1. A model training method based on cloud architecture, characterized in that: include: Determine the data type of the target data set, the data type is used to characterize the field to which the target data set belongs and the prediction direction, the field to which the target data set belongs characterizes the type and data characteristics of the target data set, the prediction direction is the angle at which the target data set needs to be predicted, and the target data set is the data set that needs to be predicted; Determine a target optimization algorithm for the cloud architecture and a target model framework for the cloud architecture according to the data type of the target data set, and determine a reference threshold and a reference weight set according to the target optimization algorithm for the cloud architecture, wherein the reference threshold and the reference weight set are a set consisting of thresholds and weight values ​​of the target model framework for the cloud architecture; Determine a target feature set according to a target data set, wherein the target feature set is a set of common features selected from different feature sets of the target data set; Determine a target threshold and a target weight set according to the target model framework, reference threshold and reference weight set of the cloud architecture, and fuse the target model framework, target threshold and target weight set of the cloud architecture to obtain a reference model; The reference model is trained according to the target feature set and a target model is determined.

2. The method according to claim 1, characterized in that The method of determining the target optimization algorithm of the cloud architecture and the target model framework of the cloud architecture according to the data type of the target data set includes: Matching corresponding historical case data from the knowledge base cloud according to the data features of the target data set, wherein the historical case data includes: a model framework and an optimization algorithm corresponding to the historical data set, and the historical data set and the target data set are of the same type; Matching a target model framework of the cloud architecture corresponding to the target data set from a model library of the cloud architecture according to the historical case data; A target optimization algorithm for the cloud architecture is matched from an optimization algorithm library for the cloud architecture according to the historical case data.

3. The method according to claim 1, characterized in that Determining a reference threshold and a reference weight set according to the target optimization algorithm of the cloud architecture includes: Determine the network structure of the target model framework of the cloud architecture; Determine the objective function of the target optimization algorithm of the cloud architecture according to the network structure of the target model framework of the cloud architecture; The reference thresholds and reference weight sets of the target model frameworks of different cloud architectures are determined according to the objective function.

4. The method according to claim 1, characterized in that Determining the target feature set according to the target data set includes: Select multiple target feature selection algorithms from the feature selection algorithm library of the cloud architecture according to the data features of the target data set; Extracting features from the target data set according to the multiple target feature selection algorithms to obtain multiple reference feature sets; Common features are screened out from the multiple reference feature sets, and the common features are combined into a target feature set.

5. The method according to claim 1, characterized in that The step of determining a target threshold and a target weight set according to the target model framework, the reference threshold and the reference weight set of the cloud architecture, and fusing the target model framework, the target threshold and the target weight set of the cloud architecture to obtain a reference model includes: Matching a corresponding target threshold and target weight set from a reference threshold and a reference weight set according to a target model framework of the cloud architecture; The target model framework, target threshold and target weight set of the cloud architecture are matched to obtain a reference model.

6. The method according to claim 1, characterized in that The step of training the reference model and determining the target model according to the target feature set includes: Taking the target feature set as input of a reference model, performing prediction through the reference model, and obtaining a reference prediction result; A target model is determined according to the reference prediction result and the target data set.

7. The method according to claim 6, characterized in that The determining the target model according to the reference prediction result and the target data set includes: Determining an error index based on the reference prediction result and the target data set; The error indicators are sorted, and the reference model corresponding to the minimum error indicator is used as the target model.

8. A model training device based on cloud architecture, characterized in that: include: A data type determination module is used to determine the data type of a target data set, wherein the data type is used to characterize the field to which the target data set belongs and the prediction direction, wherein the field to which the target data set belongs characterizes the type and data characteristics of the target data set, wherein the prediction direction is the angle at which the target data set needs to be predicted, and the target data set is the data set that needs to be predicted; A framework determination module is used to determine a target optimization algorithm of a cloud architecture and a target model framework of the cloud architecture according to a data type of a target data set, and to determine a reference threshold and a reference weight set according to the target optimization algorithm of the cloud architecture, wherein the reference threshold and the reference weight set are a set consisting of a threshold and a weight value of the target model framework of the cloud architecture; A feature set determination module, used to determine a target feature set according to a target data set, wherein the target feature set is a set of common features selected from different feature sets of the target data set; A reference model determination module, used to determine a target threshold and a target weight set according to the target model framework, reference threshold and reference weight set of the cloud architecture, and to fuse the target model framework, target threshold and target weight set of the cloud architecture to obtain a reference model; The target model determination module is used to train the reference model and determine the target model according to the target feature set.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the cloud-based architecture model training method described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the cloud-based architecture model training method described in any one of claims 1-7 when executed.