Training deployment method and device based on kylin operating system

By adopting a Kirin operating system-based method in the training and deployment system, using the decision tree and regression tree model to process historical training data, generate training and deployment models, the problem of data security risks in foreign systems is solved, and efficient and accurate domestic production is achieved. Training plan deployment.

CN118942299APending Publication Date: 2024-11-12BEIJING SPARK SPOT TECH CO LTD

Patent Information

Application Number
CN202410992031.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-23
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The existing training and deployment systems are mainly developed by foreign companies, which pose data security risks, affect the technological development and industrial security of domestic enterprises, and lack domestic solutions.

Method used

Based on the Kirin operating system, by collecting historical training data, extracting balanced training features, inputting decision tree and regression tree model, generating training deployment models, and adjusting training schemes in real time, improving the efficiency and accuracy of the training deployment center.

Benefits of technology

Real-time training solution deployment based on artificial intelligence is realized, the efficiency and accuracy of the training deployment center is improved, the dependence on foreign technologies is reduced, and data security is enhanced.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure HDA0004959132340000011
    Figure HDA0004959132340000011
  • Figure HDA0004959132340000021
    Figure HDA0004959132340000021
  • Figure HDA0004959132340000022
    Figure HDA0004959132340000022
Patent Text Reader

Abstract

The embodiment of the invention provides a training deployment method and device based on a kylin operating system. The method comprises the steps that corresponding balance training features are determined by collecting historical training data; inputting the balance training features into a feature selection layer of a preset decision tree model, determining decision tree recursion features, and determining decision tree prediction features output by the corresponding preset decision tree model according to the decision tree recursion features; inputting the decision tree prediction features into a feature integration layer of a preset regression tree model, determining training data prediction features, and determining a corresponding training deployment model according to the training data prediction features, preset real features and a back propagation algorithm; and receiving real training data, determining a corresponding training scheme according to the real training data and the training deployment model, and displaying the training scheme on a training page, so that a trainer performs training according to the training scheme, and the method can perform real-time training scheme deployment based on artificial intelligence, and improves the efficiency and accuracy of a training deployment center.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of data processing, and specifically to a training deployment method and device based on the Kylin operating system. Background Art

[0002] The training deployment system is mainly used to monitor and manage the process of exercises and training. The system has real-time data collection and distribution capabilities, and can provide leaders and instructors with accurate and timely training evaluation and feedback, thereby helping to improve training results.

[0003] The training deployment center is the core function of the training deployment system. It is responsible for selecting and configuring training subjects and starting training, which is equivalent to the instructor in the actual training process. Most of the existing training deployment center (instructor) modules pre-set some basic rules, including training objectives, training difficulty, training cycle, and resource allocation, and select corresponding training and exercise plans based on these basic rules.

[0004] At present, the mainstream training deployment systems on the market are mainly developed by foreign companies, and there is no domestic training deployment system. Although the use of foreign training deployment systems can realize intelligent training deployment in terms of function, there are also data security risks. Especially in the context of the increasingly complex global trade environment, dependence on foreign technology may affect the technological development and industrial security of domestic enterprises. Therefore, there is an urgent need for a method that can replace foreign training deployment, realize real-time collection of training data, and continuously adjust and optimize subject configuration. Summary of the invention

[0005] In response to the problems in the prior art, the present application provides a training deployment method and device based on the Kylin operating system, which can deploy real-time training plans based on artificial intelligence and improve the efficiency and accuracy of the training deployment center.

[0006] In order to solve at least one of the above problems, the present application provides the following technical solutions:

[0007] In a first aspect, the present application provides a training deployment method based on the Kylin operating system, comprising:

[0008] Collect historical training data, perform feature extraction operations on the historical training data according to preset feature extraction rules, determine corresponding training features, and determine corresponding balanced training features according to the training features and a preset feature balancing algorithm;

[0009] Input the balanced training features into the feature selection layer of a preset decision tree model, determine the corresponding decision tree selection features according to a preset segmentation rule, input the decision tree selection features into the splitting layer of the preset decision tree model for recursive splitting operation, determine the corresponding decision tree recursive features, and determine the corresponding decision tree prediction features output by the preset decision tree model according to the decision tree recursive features;

[0010] Input the decision tree prediction feature into the feature integration layer of the preset regression tree model, determine the corresponding training data prediction feature according to the decision tree prediction feature, the regression tree prediction feature output by the pruning layer in the preset regression tree model, and the preset model integration algorithm, and determine the corresponding training deployment model according to the training data prediction feature, the preset real feature, and the back propagation algorithm;

[0011] Receive real training data, determine a corresponding training plan based on the real training data and the training deployment model, and display the training plan on a training page so that the trainer can perform training according to the training plan.

[0012] Furthermore, the determining corresponding balanced training features according to the training features and a preset feature balancing algorithm includes:

[0013] Performing a minority class recognition operation on the training features according to a preset minority class recognition rule, and determining corresponding nearest neighbor training features according to the minority class training features obtained after the minority class recognition operation and a preset K nearest neighbor algorithm;

[0014] A corresponding balanced training feature is determined according to the minority class training feature and the nearest neighbor training feature.

[0015] Furthermore, the determining corresponding balanced training features according to the minority class training features and the nearest neighbor training features includes:

[0016] Performing a distance calculation operation on the minority class training feature and the nearest neighbor training feature to determine a corresponding training feature distance;

[0017] A corresponding synthetic training feature is determined according to the training feature distance and a preset ratio rule, and a feature adding operation is performed on the training feature according to the synthetic training feature to determine a corresponding balanced training feature.

[0018] Furthermore, the step of inputting the balanced training features into a feature selection layer of a preset decision tree model and determining corresponding decision tree selection features according to a preset segmentation rule includes:

[0019] Input the balanced training features into the feature selection layer of a preset decision tree model, and determine the corresponding feature segmentation points according to the preset Gini impurity;

[0020] Determine corresponding decision tree selection features according to the feature segmentation points and the balanced training features.

[0021] Furthermore, the step of inputting the decision tree selection feature into a splitting layer of a preset decision tree model to perform a recursive splitting operation, determining a corresponding decision tree recursive feature, and determining a decision tree prediction feature output by a corresponding preset decision tree model according to the decision tree recursive feature comprises:

[0022] Inputting the decision tree selection features into the splitting layer of a preset decision tree model, performing a recursive splitting operation according to a preset maximum tree depth, and determining corresponding decision tree recursive features;

[0023] The decision tree recursive features are pruned to determine the decision tree prediction features output by the corresponding preset decision tree model, wherein the decision tree prediction features are used to predict category labels.

[0024] Furthermore, before inputting the decision tree prediction features into the feature integration layer of the preset regression tree model, the method includes:

[0025] Inputting the balanced training features into a feature selection layer of a preset regression tree model, performing a feature selection operation according to a preset mean square error algorithm to determine corresponding regression tree selection features, performing a recursive splitting operation according to the regression tree selection features to determine corresponding regression tree recursive features;

[0026] The regression tree recursive features are input into the pruning layer of the preset regression tree model, the regression tree recursive features are pruned, the corresponding regression tree prediction features are determined, and the regression tree prediction features are input into the feature integration layer of the preset regression tree model, wherein the regression tree prediction features are used to predict continuous values.

[0027] Furthermore, the determining of corresponding training data prediction features according to the decision tree prediction features, the regression tree prediction features output by the pruning layer in the preset regression tree model, and the preset model integration algorithm includes:

[0028] Perform a weight matching operation based on the decision tree prediction feature and the regression tree prediction feature output by the pruning layer in the preset regression tree model to determine the corresponding decision tree prediction feature weight and regression tree prediction feature weight;

[0029] A feature weighted averaging operation is performed according to the decision tree prediction feature weights and the regression tree prediction feature weights to determine the corresponding training data prediction features.

[0030] In a second aspect, the present application provides a training deployment device based on a Kylin operating system, comprising:

[0031] A data preprocessing module, used to collect historical training data, extract training features from the historical training data according to a preset feature extraction rule, and determine corresponding balanced training features according to the training features and a preset feature balancing algorithm;

[0032] A decision tree feature prediction module, used to input the balanced training feature into a feature selection layer of a preset decision tree model, determine the corresponding decision tree selection feature according to a preset segmentation rule, input the decision tree selection feature into a splitting layer of a preset decision tree model for recursive splitting operation, determine the corresponding decision tree recursive feature, and determine the corresponding decision tree prediction feature output by the preset decision tree model according to the decision tree recursive feature;

[0033] A training deployment model determination module is used to input the decision tree prediction feature into the feature integration layer of the preset regression tree model, determine the corresponding training data prediction feature according to the decision tree prediction feature, the regression tree prediction feature output by the pruning layer in the preset regression tree model, and the preset model integration algorithm, and determine the corresponding training deployment model according to the training data prediction feature, the preset real feature, and the back propagation algorithm;

[0034] The training program display module is used to receive real training data, determine the corresponding training program according to the real training data and the training deployment model, and display the training program on the training page so that the trainer can perform training according to the training program.

[0035] In a third aspect, the present application provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the training deployment method based on the Kylin operating system when executing the program.

[0036] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the training deployment method based on the Kylin operating system.

[0037] In a fifth aspect, the present application provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the training deployment method based on the Kylin operating system.

[0038] It can be seen from the above technical scheme that the present application provides a training deployment method and device based on the Kylin operating system, which determines the corresponding balanced training features by collecting historical training data; inputs the balanced training features into the feature selection layer of the preset decision tree model to determine the decision tree recursive features, and determines the corresponding decision tree prediction features output by the preset decision tree model according to the decision tree recursive features; inputs the decision tree prediction features into the feature integration layer of the preset regression tree model to determine the training data prediction features, and determines the corresponding training deployment model according to the training data prediction features, preset real features and the back propagation algorithm; receives real training data, determines the corresponding training plan according to the real training data and the training deployment model, and displays the training plan on the training page so that the trainer can train according to the training plan, thereby enabling the deployment of real-time training plans based on artificial intelligence to improve the efficiency and accuracy of the training deployment center. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0040] Figure 1 This is one of the flow charts of the training deployment method based on the Kylin operating system in the embodiment of the present application;

[0041] Figure 2 This is the second flow chart of the training deployment method based on the Kylin operating system in the embodiment of the present application;

[0042] Figure 3 This is the third flow chart of the training deployment method based on the Kylin operating system in the embodiment of the present application;

[0043] Figure 4 This is a fourth flowchart of the training deployment method based on the Kylin operating system in an embodiment of the present application;

[0044] Figure 5 This is the fifth flow chart of the training deployment method based on the Kylin operating system in the embodiment of the present application;

[0045] Figure 6 This is the sixth flow chart of the training deployment method based on the Kylin operating system in the embodiment of the present application;

[0046] Figure 7 This is the seventh flow chart of the training deployment method based on the Kylin operating system in the embodiment of the present application;

[0047] Figure 8 This is a structural diagram of a training deployment device based on the Kylin operating system in an embodiment of the present application;

[0048] Fig. 9 It is a schematic diagram of the structure of an electronic device in an embodiment of the present application.

[0049] Reference numerals:

[0050] Electronic device 9600, central processing unit 9100, memory 9140, communication module 9110, input unit 9120, audio processor 9130, display 9160, power supply 9170, buffer memory 9141, application / function storage unit 9142, data storage unit 9143, driver program storage unit 9144, antenna 9111, speaker 9131, microphone 9132. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0052] The acquisition, storage, use, and processing of data in the technical solution of this application comply with the relevant provisions of national laws and regulations.

[0053] Considering that the mainstream training deployment systems on the market are mainly developed by foreign companies, and there is no domestic training deployment system, although the use of foreign training deployment systems can realize intelligent training deployment in terms of function, there are also data security risks. Especially in the context of the increasingly complex global trade environment, dependence on foreign technology may affect the technological development and industrial security of domestic enterprises. Therefore, there is an urgent need for a method that can replace foreign training deployment. The present application provides a training deployment method and device based on the Kylin operating system, which determines the corresponding balanced training features by collecting historical training data; inputs the balanced training features into the feature selection layer of the preset decision tree model, determines the decision tree recursive features, and determines the corresponding decision tree prediction features output by the preset decision tree model according to the decision tree recursive features; inputs the decision tree prediction features into the feature integration layer of the preset regression tree model, determines the training data prediction features, and determines the corresponding training deployment model according to the training data prediction features, the preset real features and the back propagation algorithm; receives the real training data, determines the corresponding training plan according to the real training data and the training deployment model, and displays the training plan on the training page, so that the trainer can train according to the training plan, thereby being able to deploy the real-time training plan based on artificial intelligence, and improving the efficiency and accuracy of the training deployment center.

[0054] In order to deploy real-time training solutions based on artificial intelligence and improve the efficiency and accuracy of the training deployment center, this application provides an embodiment of a training deployment method based on the Kylin operating system, see Figure 1 The training deployment method based on the Kylin operating system specifically includes the following contents:

[0055] Step S101: collecting historical training data, performing feature extraction operations on the historical training data according to preset feature extraction rules, determining corresponding training features, and determining corresponding balanced training features according to the training features and a preset feature balancing algorithm;

[0056] Optionally, in this embodiment, the purpose of this step is to preprocess the collected historical data to lay a solid data foundation for the subsequent training and deployment of the model.

[0057] Optionally, in this embodiment, the technical means adopted in this step is to perform data enhancement on the extracted training data features, mainly involving SMOTE (Synthetic Minority Over-sampling Technique) to balance the class-unbalanced data set, and improve the performance of the model on the minority class by synthesizing the minority class data.

[0058] Optionally, in this embodiment, the preset feature extraction rule is to perform data cleaning on historical data, process missing values ​​and outliers therein, extract numerical features, category features, time features, etc. from the cleaned data according to specific training requirements and data characteristics, and standardize these extracted features to obtain training features.

[0059] Optionally, in this embodiment, the preset feature balancing algorithm uses the SMOTE algorithm to analyze the category distribution in the training features, confirm whether there is category imbalance, and generate synthetic samples for minority category data to balance the data set. Apply the same feature extraction rules to the generated synthetic samples, re-extract features, and generate balanced training features.

[0060] Optionally, in this embodiment, the SMOTE algorithm first identifies the minority class in the data set and selects a sample point from the minority class. Secondly, the nearest neighbor point of the sample point is found through the K nearest neighbor algorithm, and the distance between the minority class sample point and its nearest neighbor point is calculated. Then, according to the preset proportional rule (distance <1), a new sample point is generated on the distance, and the new sample point is added to the original data to obtain balanced data.

[0061] Optionally, in this embodiment, for example, in a simulated training scenario, the comprehensive ability of most trainees fluctuates within a specific range, which will cause the collected training data to be concentrated within a specific range. For example, the time interval from when a trainee receives an instruction to when he completes the instruction is about 3 seconds, and the collected training set data is very likely to be concentrated between {2.85s-3.25s}. Some trainees with outstanding abilities will complete the task in 1s-2s, and some trainees who are not good at completing such tasks will complete the task in 4s-5s.

[0062] This will lead to a problem. More than 90% of the collected data sets will be concentrated in the range of {2.85s-3.25s}, with a few categories appearing in very rare cases. Such data sets, due to the lack of excellent ability data and unqualified ability data, will cause the model to tend to predict average results during training, resulting in the training plan generated for excellent trainers not being targeted. Trainers cannot improve according to this plan, and coaches cannot better identify trainers with talent in this training and conduct targeted training deployment.

[0063] For example, a model trained using balanced training features can do the following:

[0064] For trainers with lower skill levels, basic skills training and tactical decision-making training are recommended.

[0065] For trainers with higher skill levels, advanced tactical training and complex battlefield simulations are recommended.

[0066] It can be understood that in this embodiment, the SMOTE algorithm is applied to balance the features of the data set, which increases the number of samples in the minority categories and makes the category distribution in the training data more balanced, so that the model has more opportunities to learn the features of the trainers that can adapt well to the training items and the features of the trainers that cannot adapt to the training items, thereby improving the ability to make targeted predictions for the minority categories.

[0067] It can be understood that in this embodiment, by generating synthetic samples, the model will not be overly dependent on concentrated common samples, reducing the risk of overfitting and being able to better evaluate its effect in practical applications.

[0068] Step S102: input the balanced training features into the feature selection layer of the preset decision tree model, determine the corresponding decision tree selection features according to the preset segmentation rules, input the decision tree selection features into the splitting layer of the preset decision tree model for recursive splitting operation, determine the corresponding decision tree recursive features, and determine the corresponding decision tree prediction features output by the preset decision tree model according to the decision tree recursive features;

[0069] Optionally, in this embodiment, the purpose of this step is to obtain a decision tree prediction feature, which is used to predict category labels and is mainly used for strategic guidance. For example, an instructor decides on a training subject.

[0070] Optionally, in this embodiment, the technical means adopted in this step is to input the balanced training features obtained in step S101 into the decision tree model for model training to obtain the decision tree prediction features.

[0071] Optionally, in this embodiment, the preset segmentation rule is Gini impurity, which is a method used in the decision tree algorithm to measure the purity of a data set and is used for feature selection. It is based on probability and calculates the uneven distribution of each category label in the data set.

[0072] Optionally, in this embodiment, the Gini impurity algorithm calculates the Gini impurity of the data set for each possible segmentation point of each feature, and selects the segmentation point that reduces the Gini impurity the most. This segmentation point divides the data set into two parts, so that the sum of the Gini impurities of the two parts is minimized.

[0073] The balanced training features are divided into two selection features according to the position of the split point on the balanced training features. The selection feature determines how the current node of the decision tree will be split. This feature will define a rule of the decision tree, for example: "If feature A is greater than a certain threshold, then go to the left branch; otherwise, go to the right branch."

[0074] Optionally, in this embodiment, the recursive splitting operation is to input the selected selection features into the splitting layer. According to the feature selection and splitting point, the data is divided into two or more subsets, and each subset enters the next layer of the tree. The feature selection and splitting process is repeated recursively for each subset. Each split will divide the data set into two parts according to the selected features, and then the splitting process is repeated on these two parts of the data, and new recursive features are determined each time the split. These features are the optimal features of the current splitting node and are used to further split the data set until the stopping condition is met. In this process, the nodes except the root node represent a series of decisions that have been made. The end node of the tree, without further splitting, will contain a category label, which is the recursive feature of the decision tree.

[0075] Among them, the stopping condition is the maximum tree depth condition, that is, a maximum tree depth is preset, and when the decision tree model splits to the preset depth, the splitting stops. It can be understood that the maximum tree depth is a hyperparameter of a model set according to specific business and training needs. Setting the maximum tree depth helps the model to better predict features.

[0076] Optionally, in this embodiment, pruning is a technique to reduce the complexity of the decision tree to prevent overfitting. The pruning operation in this embodiment adopts a post-pruning operation, which first allows the tree to grow completely to obtain the recursive features of the decision tree, and then prunes those branches that do not contribute much to the prediction from the bottom.

[0077] It is understandable that the pruning operation evaluates the split effect of each node and removes those splits that cannot significantly reduce the impurity. After the pruning operation, the decision tree model will generate the final decision tree prediction features. The decision tree prediction features will be used to predict the category labels of new data. The prediction features of each node define the path from the root node to the node, and this path corresponds to a specific prediction rule.

[0078] It can be understood that through the above steps, the decision tree model can learn prediction rules from the training data and can make effective predictions on new data.

[0079] Step S103: input the decision tree prediction feature into the feature integration layer of the preset regression tree model, determine the corresponding training data prediction feature according to the decision tree prediction feature, the regression tree prediction feature output by the pruning layer in the preset regression tree model, and the preset model integration algorithm, and determine the corresponding training deployment model according to the training data prediction feature, the preset real feature, and the back propagation algorithm;

[0080] Optionally, in this embodiment, the purpose of this step is to obtain a trained training deployment model so that instructors and teachers can obtain targeted prediction solutions and suggestions in real simulated training scenarios.

[0081] Optionally, in this embodiment, the technical means used in this step is to input the decision tree prediction features obtained in the above step S102 into the feature integration layer of the regression tree model to integrate the decision tree prediction features and the regression tree prediction features, wherein the regression tree prediction features are obtained by inputting the balanced training features obtained in step S101 into the preset regression tree model. The decision tree model and the regression tree model are trained synchronously, but the decision tree prediction features obtained by training are integrated in the additional feature integration layer of the regression tree model to obtain integrated features.

[0082] It can be understood that the decision tree prediction features are used for the prediction of classification problems and strategic-level guidance, for example, predicting appropriate training programs for instructors; the regression tree prediction features are used for the prediction of continuous values ​​and tactical-level guidance, for example, predicting the reaction time or score in a training environment for trainers, which can evaluate the qualifications of trainers and reasonably allocate training resources.

[0083] Optionally, in this embodiment, the regression tree prediction feature is the output of the predicted regression tree model. The preset regression tree model is also a decision tree. Unlike the decision tree model, it is used to predict continuous numerical data rather than discrete labels in classification problems. Among them, the leaf nodes of the decision tree contain category labels, indicating the predicted category. The leaf nodes of the regression tree contain numerical values, indicating the predicted continuous output.

[0084] Optionally, in this embodiment, before inputting the decision tree prediction features and the regression tree prediction features into the feature integration layer of the regression tree model, the balanced training features obtained in step S101 are input into the regression tree model for model training to obtain the regression tree prediction features.

[0085] It can be understood that by using the balanced training features obtained in step S101 to simultaneously train the regression tree model and the decision tree model, and inputting the obtained regression tree prediction features and decision tree prediction features into the additional feature integration layer of the regression tree model for feature fusion, an integrated feature having both the regression tree prediction features and the decision tree prediction features can be obtained.

[0086] Optionally, in this embodiment, the balanced training features obtained in step S101 are input into the feature selection layer of the regression tree model, and feature selection is performed using the method of minimizing the mean square error (MSE) to determine which feature to use for splitting. After obtaining the regression tree selection features, the data is divided into two or more subsets according to the selected features and splitting points, and each subset enters the next layer of the tree. For each subset, the feature selection and splitting process is repeated until the stopping condition is met, and the stopping condition here is also the maximum tree depth. After the regression tree model stops splitting, a regression tree recursive feature is obtained.

[0087] Optionally, in this embodiment, the regression tree recursive features are pruned to cut off some branches to prevent the model from overfitting, which is used to reduce the complexity of the model and improve the generalization ability. The pruned regression tree model will generate the final regression tree prediction features. The regression tree prediction features will be used to predict the values ​​of new data. The prediction features of each node define the path from the root node to the node, and this path corresponds to a specific prediction rule.

[0088] It can be understood that through the above steps, the regression tree model can learn prediction rules from the training data and can make effective predictions on new data.

[0089] Optionally, in this embodiment, after obtaining the decision tree prediction feature and the regression tree prediction feature, the two features are fused in an additional feature fusion layer of the regression tree model through a model-to-model integration algorithm.

[0090] Specifically, first, weights are assigned according to the confidence of the model prediction. The higher the confidence, the greater the model weight. The decision tree prediction feature weights and the regression tree prediction feature weights are determined according to the matching results.

[0091] Secondly, the prediction features of the decision tree and regression tree are weighted averaged with the determined weights. The specific operation is to multiply the prediction features of each model by its weight and then sum them up:

[0092] Final prediction features = W1*decision tree prediction features + W2*regression tree prediction features, where W1 and W2 are the weights of the decision tree and regression tree prediction features, respectively. The final generated training data prediction features are based on the weighted average features, which combine the prediction capabilities of the two models to improve the accuracy and stability of the overall prediction.

[0093] It can be understood that the above steps combine the advantages of different models, reduce the bias and overfitting problems that may exist in a single model, and thus improve the overall performance of the prediction model.

[0094] Step S104: receiving real training data, determining a corresponding training plan according to the real training data and the training deployment model, and displaying the training plan on a training page so that the trainer can perform training according to the training plan.

[0095] Optionally, in this embodiment, real training data is received, and data prediction is performed through a trained model to obtain a training plan that best suits the current state of the trainer, so that the trainer can perform effective training improvements, and at the same time can help the instructor provide strategic guidance.

[0096] Optionally, in this embodiment, an example is given:

[0097] Suppose a special operations team needs tactical training to improve its combat capabilities in a complex battlefield environment.

[0098] First, during the training process, operational data of team members are collected, including response time, tactical choices, task completion, etc.

[0099] Secondly, use the trained deployment model to predict classification and regression problems, including:

[0100] The decision tree part is used to classify the skill level and tactical decision type of the trainer, and provide strategic guidance to the instructor based on the skill level and tactical decision type of the trainer. If the decision tree model identifies that a trainer performs poorly in the "direct attack decision" type, the instructor can adjust the training plan and add relevant simulation training tasks.

[0101] The regression tree part is used to predict the trainer's performance in a specific task and provide tactical guidance to the trainer. The regression tree model predicts that a trainer may face a higher risk in the "quick response" task, and the trainer can conduct relevant tactical training and risk assessment in advance.

[0102] Then, during the training process, the trainee's operation data is collected in real time and fed back to the model so that the model can generate analysis results in real time based on the trainee's training status. At the same time, the instructor can adjust the training content and methods in real time based on the analysis results of the model, and the trainee can adjust his or her tactical actions based on the analysis results to improve task completion.

[0103] For example: If a trainee performs poorly in rapid response training, the instructor can increase the training time and difficulty of the trainee, and the trainee can strengthen his or her attention in rapid response exercises, conduct self-reflection and improvement.

[0104] It is understandable that, using the training deployment model trained in steps S101 to S103, the instructor can formulate a training plan more scientifically, improve the pertinence and effect of the training, and adjust the training content and methods in real time as the training progresses, thereby improving the flexibility and adaptability of the training. The trainer can conduct tactical training more effectively, improve the task completion, and timely understand his own performance and progress as the training progresses, thereby enhancing the enthusiasm for training.

[0105] This example demonstrates how this embodiment trains a training deployment model based on historical training data. This training deployment model that integrates decision tree and regression tree models can generate predictions for classification and regression problems. Instructors and trainers can conduct training more effectively, improving the scientific nature and practicality of training.

[0106] From the above description, it can be seen that the training deployment method based on the Kylin operating system provided in the embodiment of the present application can determine the corresponding balanced training features by collecting historical training data; input the balanced training features into the feature selection layer of the preset decision tree model, determine the decision tree recursive features, and determine the corresponding decision tree prediction features output by the preset decision tree model according to the decision tree recursive features; input the decision tree prediction features into the feature integration layer of the preset regression tree model, determine the training data prediction features, and determine the corresponding training deployment model according to the training data prediction features, preset real features and the back propagation algorithm; receive real training data, determine the corresponding training plan according to the real training data and the training deployment model, and display the training plan on the training page so that the trainer can train according to the training plan, thereby enabling the deployment of real-time training plans based on artificial intelligence to improve the efficiency and accuracy of the training deployment center.

[0107] In one embodiment of the training deployment method based on the Kylin operating system of the present application, see Figure 2 , and can also include the following:

[0108] Step S201: performing a minority class recognition operation on the training features according to a preset minority class recognition rule, and determining corresponding nearest neighbor training features according to the minority class training features obtained after the minority class recognition operation and a preset K nearest neighbor algorithm;

[0109] Step S202: determining corresponding balanced training features according to the minority class training features and the nearest neighbor training features.

[0110] Optionally, in this embodiment, first, the minority class in the data set is identified and a sample point is selected from the minority class. Secondly, the nearest neighbor point of the sample point is found through the K nearest neighbor algorithm, and the distance between the minority class sample point and its nearest neighbor point is calculated. Then, according to the preset proportional rule (distance <1), a new sample point is generated on the distance, and the new sample point is added to the original data to obtain balanced data.

[0111] Optionally, in this embodiment, for example, in a simulated training scenario, the comprehensive ability of most trainees fluctuates within a specific range, which will cause the collected training data to be concentrated within a specific range. For example, the time interval from when a trainee receives an instruction to when he completes the instruction is about 3 seconds, and the collected training set data is very likely to be concentrated between {2.85s-3.25s}. Some trainees with outstanding abilities will complete the task in 1s-2s, and some trainees who are not good at completing such tasks will complete the task in 4s-5s.

[0112] This will lead to a problem. More than 90% of the collected data sets will be concentrated in the range of {2.85s-3.25s}, with a few categories appearing in very rare cases. Such data sets, due to the lack of excellent ability data and unqualified ability data, will cause the model to tend to predict average results during training, resulting in the training plan generated for excellent trainers not being targeted. Trainers cannot improve according to this plan, and coaches cannot better identify trainers with talent in this training and conduct targeted training deployment.

[0113] For example, a model trained using balanced training features can do the following:

[0114] For trainers with lower skill levels, basic skills training and tactical decision-making training are recommended.

[0115] For trainers with higher skill levels, advanced tactical training and complex battlefield simulations are recommended.

[0116] It can be understood that in this embodiment, the SMOTE algorithm is applied to balance the features of the data set, which increases the number of samples in the minority categories and makes the category distribution in the training data more balanced, so that the model has more opportunities to learn the features of the trainers that can adapt well to the training items and the features of the trainers that cannot adapt to the training items, thereby improving the ability to make targeted predictions for the minority categories.

[0117] It can be understood that in this embodiment, by generating synthetic samples, the model will not be overly dependent on concentrated common samples, reducing the risk of overfitting and being able to better evaluate its effect in practical applications.

[0118] Through step S202, this embodiment obtains balanced training features, performs data balancing on the categories in the training data, and lays a solid data preprocessing foundation for subsequent model training.

[0119] In one embodiment of the training deployment method based on the Kylin operating system of the present application, see Figure 3 , and can also include the following:

[0120] Step S301: performing a distance calculation operation on the minority class training feature and the nearest neighbor training feature to determine the corresponding training feature distance;

[0121] Step S302: determining a corresponding synthetic training feature according to the training feature distance and a preset ratio rule, performing a feature adding operation on the training feature according to the synthetic training feature, and determining a corresponding balanced training feature.

[0122] Optionally, in this embodiment, the distance between the minority class sample point and its nearest neighbor point is calculated, and then, according to a preset proportional rule (distance <1), a new sample point (i.e., a synthetic training feature) is generated on the distance, and the new sample point is added to the original data to obtain balanced data.

[0123] Optionally, in this embodiment, the distance calculation selects an appropriate distance measurement method, such as Euclidean distance.

[0124] Optionally, in this embodiment, the ratio rule is the basis for determining the number of new sample points to be generated. For example:

[0125] If the distance is less than 1 (or other threshold), a new sample point is generated on the line between the minority class sample point and its nearest neighbor point. The generation of these new sample points can increase the number of samples in the minority class, thereby reducing class imbalance.

[0126] Optionally, in this embodiment, the generated new sample points are added to the original data set to obtain a more balanced data set.

[0127] Through step S302, this embodiment obtains balanced training features, performs data balancing on the categories in the training data, and lays a solid data preprocessing foundation for subsequent model training.

[0128] In one embodiment of the training deployment method based on the Kylin operating system of the present application, see Figure 4 , and can also include the following:

[0129] Step S401: inputting the balanced training features into the feature selection layer of a preset decision tree model, and determining corresponding feature segmentation points according to a preset Gini impurity;

[0130] Step S402: determining corresponding decision tree selection features according to the feature segmentation points and the balanced training features.

[0131] Optionally, in this embodiment, the Gini impurity algorithm calculates the Gini impurity of the data set for each possible segmentation point of each feature, and selects the segmentation point that reduces the Gini impurity the most. This segmentation point divides the data set into two parts, so that the sum of the Gini impurities of the two parts is minimized.

[0132] The balanced training features are divided into two selection features according to the position of the split point on the balanced training features. The selection feature determines how the current node of the decision tree will be split. This feature will define a rule of the decision tree, for example: "If feature A is greater than a certain threshold, then go to the left branch; otherwise, go to the right branch."

[0133] Through step S402, this embodiment successfully obtains the selection features according to the Gini impurity, laying the foundation for the subsequent recursive splitting in the decision tree model.

[0134] In one embodiment of the training deployment method based on the Kylin operating system of the present application, see Figure 5 , and can also include the following:

[0135] Step S501: inputting the decision tree selection feature into the splitting layer of a preset decision tree model, performing a recursive splitting operation according to a preset maximum tree depth, and determining the corresponding decision tree recursive feature;

[0136] Step S502: performing a pruning operation on the decision tree recursive features to determine the decision tree prediction features output by the corresponding preset decision tree model, wherein the decision tree prediction features are used to predict category labels.

[0137] Optionally, in this embodiment, the recursive splitting operation is to input the selected selection features into the splitting layer. According to the feature selection and splitting point, the data is divided into two or more subsets, and each subset enters the next layer of the tree. The feature selection and splitting process is repeated recursively for each subset. Each split will divide the data set into two parts according to the selected features, and then the splitting process is repeated on these two parts of the data, and new recursive features are determined each time the split. These features are the optimal features of the current splitting node and are used to further split the data set until the stopping condition is met. In this process, the nodes except the root node represent a series of decisions that have been made. The end node of the tree, without further splitting, will contain a category label, which is the recursive feature of the decision tree.

[0138] Among them, the stopping condition is the maximum tree depth condition, that is, a maximum tree depth is preset, and when the decision tree model splits to the preset depth, the splitting stops. It can be understood that the maximum tree depth is a hyperparameter of a model set according to specific business and training needs. Setting the maximum tree depth helps the model to better predict features.

[0139] Optionally, in this embodiment, pruning is a technique to reduce the complexity of the decision tree to prevent overfitting. The pruning operation in this embodiment adopts a post-pruning operation, which first allows the tree to grow completely to obtain the recursive features of the decision tree, and then prunes those branches that do not contribute much to the prediction from the bottom.

[0140] It is understandable that the pruning operation evaluates the split effect of each node and removes those splits that cannot significantly reduce the impurity. After the pruning operation, the decision tree model will generate the final decision tree prediction features. The decision tree prediction features will be used to predict the category labels of new data. The prediction features of each node define the path from the root node to the node, and this path corresponds to a specific prediction rule.

[0141] Through step S502, this embodiment successfully obtains the decision tree prediction features output by the decision tree model, laying a foundation for subsequent fusion with the regression tree prediction features to obtain a training deployment model.

[0142] In one embodiment of the training deployment method based on the Kylin operating system of the present application, see Figure 6 , and can also include the following:

[0143] Step S601: input the balanced training features into the feature selection layer of the preset regression tree model, perform feature selection operation according to the preset mean square error algorithm, determine the corresponding regression tree selection features, perform recursive splitting operation according to the regression tree selection features, and determine the corresponding regression tree recursive features;

[0144] Step S602: input the regression tree recursive features into the pruning layer of the preset regression tree model, perform pruning operations on the regression tree recursive features, determine corresponding regression tree prediction features, and input the regression tree prediction features into the feature integration layer of the preset regression tree model, wherein the regression tree prediction features are used to predict continuous values.

[0145] Optionally, in this embodiment, the balanced training features obtained in step S101 are input into the feature selection layer of the regression tree model, and feature selection is performed using the method of minimizing the mean square error (MSE) to determine which feature to use for splitting. After obtaining the regression tree selection features, the data is divided into two or more subsets according to the selected features and splitting points, and each subset enters the next layer of the tree. For each subset, the feature selection and splitting process is repeated until the stopping condition is met, and the stopping condition here is also the maximum tree depth. After the regression tree model stops splitting, a regression tree recursive feature is obtained.

[0146] Optionally, in this embodiment, the regression tree recursive features are pruned to cut off some branches to prevent the model from overfitting, which is used to reduce the complexity of the model and improve the generalization ability. The pruned regression tree model will generate the final regression tree prediction features. The regression tree prediction features will be used to predict the values ​​of new data. The prediction features of each node define the path from the root node to the node, and this path corresponds to a specific prediction rule.

[0147] It can be understood that through the above steps, the regression tree model can learn prediction rules from the training data and can make effective predictions on new data.

[0148] Through step S602, this embodiment successfully obtains the regression tree prediction features, laying a foundation for subsequent fusion with the decision tree prediction features to obtain a training deployment model.

[0149] In one embodiment of the training deployment method based on the Kylin operating system of the present application, see Figure 7 , and can also include the following:

[0150] Step S701: performing a weight matching operation according to the decision tree prediction feature and the regression tree prediction feature output by the pruning layer in the preset regression tree model to determine the corresponding decision tree prediction feature weight and regression tree prediction feature weight;

[0151] Step S702: performing a feature weighted average operation according to the decision tree prediction feature weight and the regression tree prediction feature weight to determine the corresponding training data prediction feature.

[0152] Optionally, in this embodiment, after obtaining the decision tree prediction feature and the regression tree prediction feature, the two features are fused in an additional feature fusion layer of the regression tree model through a model-to-model integration algorithm.

[0153] Specifically, first, weights are assigned according to the confidence of the model prediction. The higher the confidence, the greater the model weight. The decision tree prediction feature weights and the regression tree prediction feature weights are determined according to the matching results.

[0154] Secondly, the prediction features of the decision tree and regression tree are weighted averaged with the determined weights. The specific operation is to multiply the prediction features of each model by its weight and then sum them up:

[0155] Final prediction features = W1*decision tree prediction features + W2*regression tree prediction features, where W1 and W2 are the weights of the decision tree and regression tree prediction features, respectively. The final generated training data prediction features are based on the weighted average features, which combine the prediction capabilities of the two models to improve the accuracy and stability of the overall prediction.

[0156] It can be understood that the above steps combine the advantages of different models, reduce the bias and overfitting problems that may exist in a single model, and thus improve the overall performance of the prediction model.

[0157] Through step S702, this embodiment successfully fuses the decision tree prediction features and the regression tree prediction features to obtain the fused training data prediction features, which combine the prediction capabilities of the two models to improve the accuracy and stability of the overall prediction.

[0158] In order to deploy real-time training solutions based on artificial intelligence and improve the efficiency and accuracy of the training deployment center, the present application provides an embodiment of a training deployment device based on the Kylin operating system for implementing all or part of the content of the training deployment method based on the Kylin operating system, see Figure 8 The training deployment device based on the Kylin operating system specifically includes the following contents:

[0159] A data preprocessing module 10 is used to collect historical training data, extract training features from the historical training data according to a preset feature extraction rule, and determine corresponding balanced training features according to the training features and a preset feature balancing algorithm;

[0160] A decision tree feature prediction module 20 is used to input the balanced training feature into a feature selection layer of a preset decision tree model, determine the corresponding decision tree selection feature according to a preset segmentation rule, input the decision tree selection feature into a splitting layer of a preset decision tree model for recursive splitting operation, determine the corresponding decision tree recursive feature, and determine the corresponding decision tree prediction feature output by the preset decision tree model according to the decision tree recursive feature;

[0161] The training deployment model determination module 30 is used to input the decision tree prediction feature into the feature integration layer of the preset regression tree model, determine the corresponding training data prediction feature according to the decision tree prediction feature, the regression tree prediction feature output by the pruning layer in the preset regression tree model and the preset model integration algorithm, and determine the corresponding training deployment model according to the training data prediction feature, the preset real feature and the back propagation algorithm;

[0162] The training program display module 40 is used to receive real training data, determine the corresponding training program according to the real training data and the training deployment model, and display the training program on the training page so that the trainer can perform training according to the training program.

[0163] From the above description, it can be seen that the training deployment device based on the Kylin operating system provided in the embodiment of the present application can determine the corresponding balanced training features by collecting historical training data; input the balanced training features into the feature selection layer of the preset decision tree model, determine the decision tree recursive features, and determine the corresponding decision tree prediction features output by the preset decision tree model according to the decision tree recursive features; input the decision tree prediction features into the feature integration layer of the preset regression tree model, determine the training data prediction features, and determine the corresponding training deployment model according to the training data prediction features, preset real features and the back propagation algorithm; receive real training data, determine the corresponding training plan according to the real training data and the training deployment model, and display the training plan on the training page so that the trainer can train according to the training plan, thereby enabling the deployment of real-time training plans based on artificial intelligence to improve the efficiency and accuracy of the training deployment center.

[0164] From the hardware level, in order to deploy real-time training solutions based on artificial intelligence and improve the efficiency and accuracy of the training deployment center, the present application provides an embodiment of an electronic device for implementing all or part of the content of the training deployment method based on the Kylin operating system, and the electronic device specifically includes the following content:

[0165] Processor, memory, communication interface and bus; wherein the processor, memory and communication interface communicate with each other through the bus; the communication interface is used to realize the information transmission between the training deployment method based on the Kylin operating system and the core business system, user terminal and related database and other related devices; the logic controller can be a desktop computer, a tablet computer and a mobile terminal, etc., and the present embodiment is not limited thereto. In the present embodiment, the logic controller can be implemented with reference to the embodiment of the training deployment method based on the Kylin operating system and the embodiment of the training deployment method based on the Kylin operating system in the embodiment, and the contents thereof are incorporated herein, and the repeated parts are not repeated.

[0166] It is understandable that the user terminal may include a smart phone, a tablet electronic device, a network set-top box, a portable computer, a desktop computer, a personal digital assistant (PDA), a vehicle-mounted device, a smart wearable device, etc. Among them, the smart wearable device may include smart glasses, a smart watch, a smart bracelet, etc.

[0167] In practical applications, part of the training deployment method based on the Kylin operating system can be executed on the electronic device side as described above, or all operations can be completed in the client device. The specific selection can be based on the processing capability of the client device and the limitations of the user's usage scenario. This application does not limit this. If all operations are completed in the client device, the client device may also include a processor.

[0168] The client device may have a communication module (i.e., a communication unit) that can communicate with a remote server to achieve data transmission with the server. The server may include a server on the task scheduling center side, and other implementation scenarios may also include a server on an intermediate platform, such as a server on a third-party server platform that has a communication link with the task scheduling center server. The server may include a single computer device, or a server cluster consisting of multiple servers, or a server structure of a distributed device.

[0169] Fig. 9 FIG. 9 is a schematic block diagram of the system structure of the electronic device 9600 according to an embodiment of the present application. Fig. 9 As shown, the electronic device 9600 may include a central processor 9100 and a memory 9140; the memory 9140 is coupled to the central processor 9100. It is worth noting that Fig. 9 is exemplary; other types of structures may also be used to supplement or replace this structure to implement telecommunication functions or other functions.

[0170] In one embodiment, the training deployment method function based on the Kylin operating system can be integrated into the central processing unit 9100. The central processing unit 9100 can be configured to perform the following control:

[0171] Step S101: collecting historical training data, performing feature extraction operations on the historical training data according to preset feature extraction rules, determining corresponding training features, and determining corresponding balanced training features according to the training features and a preset feature balancing algorithm;

[0172] Step S102: input the balanced training features into the feature selection layer of the preset decision tree model, determine the corresponding decision tree selection features according to the preset segmentation rules, input the decision tree selection features into the splitting layer of the preset decision tree model for recursive splitting operation, determine the corresponding decision tree recursive features, and determine the corresponding decision tree prediction features output by the preset decision tree model according to the decision tree recursive features;

[0173] Step S103: input the decision tree prediction feature into the feature integration layer of the preset regression tree model, determine the corresponding training data prediction feature according to the decision tree prediction feature, the regression tree prediction feature output by the pruning layer in the preset regression tree model, and the preset model integration algorithm, and determine the corresponding training deployment model according to the training data prediction feature, the preset real feature, and the back propagation algorithm;

[0174] Step S104: receiving real training data, determining a corresponding training plan according to the real training data and the training deployment model, and displaying the training plan on a training page so that the trainer can perform training according to the training plan.

[0175] From the above description, it can be seen that the electronic device provided in the embodiment of the present application determines the corresponding balanced training features by collecting historical training data; inputs the balanced training features into the feature selection layer of the preset decision tree model to determine the decision tree recursive features, and determines the decision tree prediction features output by the corresponding preset decision tree model according to the decision tree recursive features; inputs the decision tree prediction features into the feature integration layer of the preset regression tree model to determine the training data prediction features, and determines the corresponding training deployment model according to the training data prediction features, preset real features and the back propagation algorithm; receives real training data, determines the corresponding training plan according to the real training data and the training deployment model, and displays the training plan on the training page so that the trainer can train according to the training plan, thereby enabling the deployment of real-time training plans based on artificial intelligence to improve the efficiency and accuracy of the training deployment center.

[0176] In another embodiment, the training deployment method based on the Kylin operating system can be configured separately from the central processing unit 9100. For example, the training deployment method based on the Kylin operating system can be configured as a chip connected to the central processing unit 9100, and the functions of the training deployment method based on the Kylin operating system can be realized through the control of the central processing unit.

[0177] like Fig. 9 As shown, the electronic device 9600 may also include: a communication module 9110, an input unit 9120, an audio processor 9130, a display 9160, and a power supply 9170. It is worth noting that the electronic device 9600 does not necessarily have to include Fig. 9 In addition, the electronic device 9600 may also include Fig. 9 For components not shown, reference may be made to the prior art.

[0178] like Fig. 9As shown, the central processing unit 9100 is sometimes also referred to as a controller or an operation control, and may include a microprocessor or other processor device and / or logic device. The central processing unit 9100 receives input and controls the operation of various components of the electronic device 9600.

[0179] The memory 9140 may be, for example, one or more of a cache, a flash memory, a hard drive, a removable medium, a volatile memory, a non-volatile memory or other suitable devices. The above-mentioned information related to the failure may be stored, and a program for executing the relevant information may also be stored. The CPU 9100 may execute the program stored in the memory 9140 to implement information storage or processing, etc.

[0180] The input unit 9120 provides input to the central processing unit 9100. The input unit 9120 is, for example, a key or a touch input device. The power supply 9170 is used to provide power to the electronic device 9600. The display 9160 is used to display display objects such as images and texts. The display may be, for example, an LCD display, but is not limited thereto.

[0181] The memory 9140 may be a solid-state memory, such as a read-only memory (ROM), a random access memory (RAM), a SIM card, etc. It may also be a memory that saves information even when the power is off, can be selectively erased, and is provided with more data, examples of which are sometimes referred to as EPROMs, etc. The memory 9140 may also be some other type of device. The memory 9140 includes a buffer memory 9141 (sometimes referred to as a buffer). The memory 9140 may include an application / function storage unit 9142, which is used to store application programs and function programs or processes for executing the operation of the electronic device 9600 through the central processor 9100.

[0182] The memory 9140 may also include a data storage unit 9143 for storing data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit 9144 of the memory 9140 may include various drivers for communication functions of the electronic device and / or for executing other functions of the electronic device (such as messaging applications, address book applications, etc.).

[0183] The communication module 9110 is a transmitter / receiver that sends and receives signals via the antenna 9111. The communication module 9110 (transmitter / receiver) is coupled to the central processor 9100 to provide input signals and receive output signals, which may be the same as the case of a conventional mobile communication terminal.

[0184] Based on different communication technologies, multiple communication modules 9110 may be provided in the same electronic device, such as a cellular network module, a Bluetooth module and / or a wireless LAN module, etc. The communication module (transmitter / receiver) 9110 is also coupled to a speaker 9131 and a microphone 9132 via an audio processor 9130 to provide an audio output via the speaker 9131 and receive an audio input from the microphone 9132, thereby realizing a common telecommunication function. The audio processor 9130 may include any suitable buffer, decoder, amplifier, etc. In addition, the audio processor 9130 is also coupled to the central processor 9100, so that recording can be performed on the local machine through the microphone 9132, and the sound stored on the local machine can be played through the speaker 9131.

[0185] The embodiments of the present application also provide a computer-readable storage medium capable of implementing all the steps of the training deployment method based on the Kylin operating system in the above-mentioned embodiment, where the execution subject is a server or a client. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, all the steps of the training deployment method based on the Kylin operating system in the above-mentioned embodiment are implemented. For example, when the processor executes the computer program, the following steps are implemented:

[0186] Step S101: collecting historical training data, performing feature extraction operations on the historical training data according to preset feature extraction rules, determining corresponding training features, and determining corresponding balanced training features according to the training features and a preset feature balancing algorithm;

[0187] Step S102: input the balanced training features into the feature selection layer of the preset decision tree model, determine the corresponding decision tree selection features according to the preset segmentation rules, input the decision tree selection features into the splitting layer of the preset decision tree model for recursive splitting operation, determine the corresponding decision tree recursive features, and determine the corresponding decision tree prediction features output by the preset decision tree model according to the decision tree recursive features;

[0188] Step S103: input the decision tree prediction feature into the feature integration layer of the preset regression tree model, determine the corresponding training data prediction feature according to the decision tree prediction feature, the regression tree prediction feature output by the pruning layer in the preset regression tree model, and the preset model integration algorithm, and determine the corresponding training deployment model according to the training data prediction feature, the preset real feature, and the back propagation algorithm;

[0189] Step S104: receiving real training data, determining a corresponding training plan according to the real training data and the training deployment model, and displaying the training plan on a training page so that the trainer can perform training according to the training plan.

[0190] From the above description, it can be seen that the computer-readable storage medium provided in the embodiment of the present application determines the corresponding balanced training features by collecting historical training data; inputs the balanced training features into the feature selection layer of the preset decision tree model, determines the decision tree recursive features, and determines the decision tree prediction features output by the corresponding preset decision tree model according to the decision tree recursive features; inputs the decision tree prediction features into the feature integration layer of the preset regression tree model, determines the training data prediction features, and determines the corresponding training deployment model according to the training data prediction features, preset real features and the back propagation algorithm; receives real training data, determines the corresponding training plan according to the real training data and the training deployment model, and displays the training plan on the training page so that the trainer can train according to the training plan, thereby enabling the deployment of real-time training plans based on artificial intelligence, thereby improving the efficiency and accuracy of the training deployment center.

[0191] The embodiments of the present application also provide a computer program product capable of implementing all the steps of the training deployment method based on the Kylin operating system in the above-mentioned embodiment, where the execution subject is a server or a client. When the computer program / instruction is executed by a processor, the steps of the training deployment method based on the Kylin operating system are implemented. For example, the computer program / instruction implements the following steps:

[0192] Step S101: collecting historical training data, performing feature extraction operations on the historical training data according to preset feature extraction rules, determining corresponding training features, and determining corresponding balanced training features according to the training features and a preset feature balancing algorithm;

[0193] Step S102: input the balanced training features into the feature selection layer of the preset decision tree model, determine the corresponding decision tree selection features according to the preset segmentation rules, input the decision tree selection features into the splitting layer of the preset decision tree model for recursive splitting operation, determine the corresponding decision tree recursive features, and determine the corresponding decision tree prediction features output by the preset decision tree model according to the decision tree recursive features;

[0194] Step S103: input the decision tree prediction feature into the feature integration layer of the preset regression tree model, determine the corresponding training data prediction feature according to the decision tree prediction feature, the regression tree prediction feature output by the pruning layer in the preset regression tree model, and the preset model integration algorithm, and determine the corresponding training deployment model according to the training data prediction feature, the preset real feature, and the back propagation algorithm;

[0195] Step S104: receiving real training data, determining a corresponding training plan according to the real training data and the training deployment model, and displaying the training plan on a training page so that the trainer can perform training according to the training plan.

[0196] From the above description, it can be seen that the computer program product provided in the embodiment of the present application determines the corresponding balanced training features by collecting historical training data; inputs the balanced training features into the feature selection layer of the preset decision tree model to determine the decision tree recursive features, and determines the decision tree prediction features output by the corresponding preset decision tree model according to the decision tree recursive features; inputs the decision tree prediction features into the feature integration layer of the preset regression tree model to determine the training data prediction features, and determines the corresponding training deployment model according to the training data prediction features, preset real features and the back propagation algorithm; receives real training data, determines the corresponding training plan according to the real training data and the training deployment model, and displays the training plan on the training page so that the trainer can train according to the training plan, thereby enabling the deployment of real-time training plans based on artificial intelligence, thereby improving the efficiency and accuracy of the training deployment center.

[0197] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, devices, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0198] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (apparatus), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0199] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0200] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0201] The present invention uses specific embodiments to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A training deployment method based on Kylin operating system, characterized in that: Applied to Kylin operating system, the method includes: Collect historical training data, perform feature extraction operations on the historical training data according to preset feature extraction rules, determine corresponding training features, and determine corresponding balanced training features according to the training features and a preset feature balancing algorithm; Input the balanced training features into the feature selection layer of a preset decision tree model, determine the corresponding decision tree selection features according to a preset segmentation rule, input the decision tree selection features into the splitting layer of the preset decision tree model for a recursive splitting operation, determine the corresponding decision tree recursive features, and determine the corresponding decision tree prediction features output by the preset decision tree model according to the decision tree recursive features; Input the decision tree prediction feature into the feature integration layer of the preset regression tree model, determine the corresponding training data prediction feature according to the decision tree prediction feature, the regression tree prediction feature output by the pruning layer in the preset regression tree model, and the preset model integration algorithm, and determine the corresponding training deployment model according to the training data prediction feature, the preset real feature, and the back propagation algorithm; Receive real training data, determine a corresponding training plan based on the real training data and the training deployment model, and display the training plan on a training page so that the trainer can perform training according to the training plan.

2. The training deployment method based on Kylin operating system according to claim 1, characterized in that: The determining corresponding balanced training features according to the training features and a preset feature balancing algorithm includes: Performing a minority class recognition operation on the training features according to a preset minority class recognition rule, and determining corresponding nearest neighbor training features according to the minority class training features obtained after the minority class recognition operation and a preset K nearest neighbor algorithm; A corresponding balanced training feature is determined according to the minority class training feature and the nearest neighbor training feature.

3. The training deployment method based on Kylin operating system according to claim 2 is characterized in that: The determining corresponding balanced training features according to the minority class training features and the nearest neighbor training features includes: Performing a distance calculation operation on the minority class training feature and the nearest neighbor training feature to determine a corresponding training feature distance; A corresponding synthetic training feature is determined according to the training feature distance and a preset ratio rule, and a feature adding operation is performed on the training feature according to the synthetic training feature to determine a corresponding balanced training feature.

4. The training deployment method based on Kylin operating system according to claim 1, characterized in that: The step of inputting the balanced training features into the feature selection layer of the preset decision tree model and determining corresponding decision tree selection features according to a preset segmentation rule comprises: Input the balanced training features into the feature selection layer of the preset decision tree model, and determine the corresponding feature segmentation points according to the preset Gini impurity; The corresponding decision tree selection feature is determined according to the feature segmentation point and the balanced training feature.

5. The training deployment method based on Kylin operating system according to claim 1, characterized in that: The step of inputting the decision tree selection feature into the splitting layer of the preset decision tree model to perform a recursive splitting operation, determining the corresponding decision tree recursive feature, and determining the decision tree prediction feature output by the corresponding preset decision tree model according to the decision tree recursive feature, comprises: Inputting the decision tree selection feature into the splitting layer of the preset decision tree model, performing a recursive splitting operation according to a preset maximum tree depth, and determining the corresponding decision tree recursive feature; A pruning operation is performed on the decision tree recursive features to determine the decision tree prediction features output by the corresponding preset decision tree model, wherein the decision tree prediction features are used to predict category labels.

6. The training deployment method based on Kylin operating system according to claim 1, characterized in that: Before inputting the decision tree prediction features into the feature integration layer of the preset regression tree model, the method includes: Inputting the balanced training features into a feature selection layer of a preset regression tree model, performing a feature selection operation according to a preset mean square error algorithm to determine corresponding regression tree selection features, performing a recursive splitting operation according to the regression tree selection features to determine corresponding regression tree recursive features; The regression tree recursive features are input into the pruning layer of the preset regression tree model, the regression tree recursive features are pruned, the corresponding regression tree prediction features are determined, and the regression tree prediction features are input into the feature integration layer of the preset regression tree model, wherein the regression tree prediction features are used to predict continuous values.

7. The training deployment method based on Kylin operating system according to claim 1, characterized in that: The determining of the corresponding training data prediction features according to the decision tree prediction features, the regression tree prediction features output by the pruning layer in the preset regression tree model, and the preset model integration algorithm includes: Perform a weight matching operation based on the decision tree prediction feature and the regression tree prediction feature output by the pruning layer in the preset regression tree model to determine the corresponding decision tree prediction feature weight and regression tree prediction feature weight; A feature weighted averaging operation is performed according to the decision tree prediction feature weights and the regression tree prediction feature weights to determine the corresponding training data prediction features.

8. A training deployment device based on Kylin operating system, characterized in that: The device comprises: A data preprocessing module, used to collect historical training data, extract training features from the historical training data according to a preset feature extraction rule, and determine corresponding balanced training features according to the training features and a preset feature balancing algorithm; A decision tree feature prediction module, used to input the balanced training feature into a feature selection layer of a preset decision tree model, determine the corresponding decision tree selection feature according to a preset segmentation rule, input the decision tree selection feature into a splitting layer of the preset decision tree model for recursive splitting operation, determine the corresponding decision tree recursive feature, and determine the corresponding decision tree prediction feature output by the preset decision tree model according to the decision tree recursive feature; A training deployment model determination module is used to input the decision tree prediction feature into the feature integration layer of the preset regression tree model, determine the corresponding training data prediction feature according to the decision tree prediction feature, the regression tree prediction feature output by the pruning layer in the preset regression tree model, and the preset model integration algorithm, and determine the corresponding training deployment model according to the training data prediction feature, the preset real feature, and the back propagation algorithm; The training program display module is used to receive real training data, determine the corresponding training program according to the real training data and the training deployment model, and display the training program on the training page so that the trainer can perform training according to the training program.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the training deployment method based on the Kylin operating system described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the training deployment method based on the Kylin operating system described in any one of claims 1 to 7 are implemented.

Citation Information

Patent Citations

  • Asthma diagnosis system based on decision tree and improved SMOTE algorithm

    CN112951413A

  • Firefighter fire scene behavior recognition model construction method and device

    CN114926854A

  • Power transformer fault diagnosis method, device and equipment and storage medium

    CN116956126A

  • Hidden markov model based data ranking for enhancement of classifier performance to classify imbalanced dataset

    US20230128462A1

Cited By

  • Method for predicting medical-seeking delay risk of gastric cancer patient

    CN119314681A