Mini-program plug-in processing method and device

Abstract

An embodiment of the present application provides a method and device for processing a mini-program plug-in, the method comprising: receiving a mini-program plug-in editing request sent by a user, extracting keywords and functional features from the mini-program plug-in editing request according to the mini-program plug-in editing request and a preset natural language processing algorithm, and determining corresponding plug-in function summary features according to the keywords and functional features; inputting decision tree prediction features into a feature integration layer of a preset regression tree model, determining corresponding plug-in classification 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, determining a corresponding plug-in recommendation model according to the plug-in classification features, preset real features, and a back propagation algorithm, and pushing the recommended plug-in output by the plug-in recommendation model to the user; the present application can process mini-program plug-ins more intelligently, accurately, and efficiently to improve user experience and the convenience of function expansion.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method and device for processing a small program plug-in. Background Art

[0002] In the modern digital application environment, mini programs, as a lightweight application, have been widely welcomed because of their features of being ready to use and no need to download and install. In order to improve the functions and user experience of mini programs, developers usually use various plug-ins to expand the functions of mini programs. However, how to efficiently and intelligently process and recommend mini program plug-ins has become an important issue in the current technical field. Although the existing plug-in processing methods have improved the functional scalability of mini programs to a certain extent, they still have many shortcomings in dealing with the diversity and dynamic changes of user needs.

[0003] First, existing methods for processing mini-program plug-ins usually rely on explicit requests and manual selection by users. This method meets the specific needs of users to a certain extent, but for users who lack technical background, the process of selecting and configuring plug-ins can be cumbersome and time-consuming. In addition, users often find it difficult to accurately describe the specific functions of the required plug-ins in their editing requests, which makes it difficult for the system to accurately match the corresponding plug-ins, affecting the user experience.

[0004] Secondly, the existing technology mainly relies on manual labeling and keyword matching methods to extract and analyze plug-in functional features. This method is not only inefficient, but also prone to inaccurate labeling or unreasonable keyword selection. With the increasing complexity and diversification of mini-program plug-in functions, manual labeling methods are obviously difficult to meet the needs. Even methods based on keyword matching are difficult to fully capture the functional features of plug-ins, resulting in limited accuracy and coverage of recommendation results.

[0005] In terms of model selection and optimization, existing plug-in processing methods usually use traditional machine learning models, such as decision trees and random forests. Although these models can improve the accuracy of recommendations to a certain extent, they still have performance bottlenecks when processing high-dimensional data and complex features. In particular, decision tree models are prone to overfitting or underfitting problems during feature selection and splitting, which affects the generalization ability and stability of the model. In addition, the existing models have low computational efficiency when processing large-scale data, making it difficult to respond to user needs in real time.

[0006] In addition, the existing plug-in processing system lacks effective model integration and optimization mechanisms. Although some systems have tried to introduce integrated learning methods, such as gradient boosting and random forest, the integration effect and computational efficiency of the model still need to be improved in practical applications. Especially in the case of multi-source heterogeneous data, how to effectively integrate the prediction results of different models and improve the accuracy and robustness of recommendations is an urgent problem to be solved.

[0007] Existing technologies also have deficiencies in feedback and optimization of recommendation results. Most existing systems rely on static rules and preset logic, lacking the ability to dynamically adjust and self-optimize. Due to the constant changes in user needs and market environment, fixed recommendation strategies are difficult to maintain long-term effectiveness. In addition, during the push process, existing systems often ignore user feedback on recommended content, lack effective feedback mechanisms and model adjustment methods, resulting in difficulty in continuously optimizing recommendation results and affecting user experience.

[0008] In summary, the existing methods for processing mini-program plug-ins still have many deficiencies and challenges in terms of capturing user needs, extracting functional features, selecting and optimizing models, and feedback and adjustment of recommendation results. In the face of increasingly complex user needs and market environments, a more intelligent, accurate, and efficient mini-program plug-in processing method is urgently needed to improve user experience and the convenience of functional expansion. Summary of the invention

[0009] In response to the problems in the prior art, the present application provides a method and device for processing mini-program plug-ins, which can process mini-program plug-ins more intelligently, accurately and efficiently to improve user experience and the convenience of functional expansion.

[0010] In order to solve at least one of the above problems, the present application provides the following technical solutions:

[0011] In a first aspect, the present application provides a method for processing a mini-program plug-in, comprising:

[0012] Receive a mini-program plug-in editing request sent by a user, extract keywords and functional features from the mini-program plug-in editing request according to the mini-program plug-in editing request and a preset natural language processing algorithm, and determine corresponding plug-in functional summary features according to the keywords and functional features;

[0013] Input the plug-in function summary feature into the feature selection layer of the preset decision tree model, determine the corresponding decision tree selection feature according to the preset segmentation rule, input the decision tree selection feature into the 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;

[0014] The decision tree prediction feature is input into the feature integration layer of the preset regression tree model, and the corresponding plug-in classification feature is determined 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. The corresponding plug-in recommendation model is determined according to the plug-in classification feature, the preset real feature, and the back propagation algorithm, and the recommended plug-in outputted by the plug-in recommendation model is pushed to the user.

[0015] Furthermore, the determining of the corresponding balancing function feature according to the plug-in function summary feature and the preset feature balancing algorithm includes:

[0016] Performing a minority class identification operation on the plug-in function summary feature according to a preset minority class identification rule, and determining a corresponding nearest neighbor function feature according to the minority class function feature obtained after the minority class identification operation and a preset K nearest neighbor algorithm;

[0017] A corresponding balanced functional feature is determined according to the minority class functional feature and the nearest neighbor functional feature.

[0018] Furthermore, the determining of the corresponding nearest neighbor functional features according to the minority class functional features obtained after the minority class recognition operation and a preset K nearest neighbor algorithm includes:

[0019] Performing a distance calculation operation on the minority class functional feature and the nearest neighbor functional feature to determine a corresponding functional feature distance;

[0020] The corresponding synthetic functional feature is determined according to the functional feature distance and a preset ratio rule, and a feature adding operation is performed on the functional feature according to the synthetic functional feature to determine the corresponding balanced functional feature.

[0021] Furthermore, the step of inputting the balance function feature 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:

[0022] Input the balancing function feature into the feature selection layer of a preset decision tree model, and determine the corresponding feature segmentation point according to the preset Gini impurity;

[0023] The corresponding decision tree selection feature is determined according to the feature segmentation point and the balance function feature.

[0024] 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:

[0025] 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;

[0026] 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.

[0027] Furthermore, before inputting the decision tree prediction features into the feature integration layer of the preset regression tree model, the method includes:

[0028] Inputting the balance function feature into the feature selection layer of a preset regression tree model, performing a feature selection operation according to a preset mean square error algorithm to determine a corresponding regression tree selection feature, performing a recursive splitting operation according to the regression tree selection feature to determine a corresponding regression tree recursive feature;

[0029] 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.

[0030] Furthermore, the determining of the corresponding plug-in classification 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:

[0031] 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;

[0032] A feature weighted average operation is performed according to the decision tree prediction feature weight and the regression tree prediction feature weight to determine the corresponding plug-in classification feature.

[0033] In a second aspect, the present application provides a mini-program plug-in processing device, including:

[0034] A plug-in collection module is used to collect historical mini-program access data sent by users, perform feature extraction operations on the historical mini-program access data according to preset feature extraction rules, determine corresponding plug-in function summary features, and determine corresponding balance function features according to the plug-in function summary features and a preset feature balance algorithm;

[0035] A plug-in classification module is used to input the balance function feature into the feature selection layer of the preset decision tree model, determine the corresponding decision tree selection feature according to the preset segmentation rule, input the decision tree selection feature into the 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;

[0036] A plug-in recommendation module is used to input the decision tree prediction features into the feature integration layer of the preset regression tree model, determine the corresponding plug-in classification 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, determine the corresponding mini-program push model according to the plug-in classification features, preset real features, and the back propagation algorithm, and perform the mini-program push operation according to the mini-program push model.

[0037] 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 mini-program plug-in processing method when executing the program.

[0038] 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 mini-program plug-in processing method.

[0039] In a fifth aspect, the present application provides a computer program product, including a computer program / instruction, which implements the steps of the mini-program plug-in processing method when executed by a processor.

[0040] It can be seen from the above technical scheme that the present application provides a method and device for processing mini-program plug-ins, which receives a mini-program plug-in editing request sent by a user, extracts keywords and functional features from the mini-program plug-in editing request according to the mini-program plug-in editing request and a preset natural language processing algorithm, and determines the corresponding plug-in function summary features according to the keywords and functional features; inputs the decision tree prediction features into the feature integration layer of the preset regression tree model, determines the corresponding plug-in classification 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, determines the corresponding plug-in recommendation model according to the plug-in classification features, the preset real features, and the back propagation algorithm, and pushes the recommended plug-in output by the plug-in recommendation model to the user, thereby being able to process the mini-program plug-in more intelligently, accurately, and efficiently to enhance user experience and the convenience of function expansion. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] 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.

[0042] Figure 1 This is one of the flowcharts of the method for processing a mini-program plug-in in an embodiment of the present application;

[0043] Figure 2 This is a second flowchart of the method for processing a mini-program plug-in in an embodiment of the present application;

[0044] Figure 3 This is a third flowchart of the method for processing a mini-program plug-in in an embodiment of the present application;

[0045] Figure 4 This is a fourth flowchart of the method for processing a mini-program plug-in in an embodiment of the present application;

[0046] Figure 5 This is a fifth flowchart of the method for processing a mini-program plug-in in an embodiment of the present application;

[0047] Figure 6 This is a sixth flowchart of the method for processing a mini-program plug-in in an embodiment of the present application;

[0048] Figure 7 This is the seventh flowchart of the method for processing a mini-program plug-in in an embodiment of the present application;

[0049] Figure 8 A structural diagram of a mini-program plug-in processing device in an embodiment of the present application;

[0050] Fig. 9 It is a schematic diagram of the structure of an electronic device in an embodiment of the present application.

[0051] Reference numerals:

[0052] 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

[0053] 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.

[0054] 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.

[0055] In view of the problems existing in the prior art, the present application provides a method and device for processing mini-program plug-ins, which receives a mini-program plug-in editing request sent by a user, extracts keywords and functional features from the mini-program plug-in editing request according to the mini-program plug-in editing request and a preset natural language processing algorithm, and determines the corresponding plug-in function summary features according to the keywords and functional features; inputs the decision tree prediction features into the feature integration layer of the preset regression tree model, determines the corresponding plug-in classification 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, determines the corresponding plug-in recommendation model according to the plug-in classification features, the preset real features, and the back propagation algorithm, and pushes the recommended plug-in output by the plug-in recommendation model to the user, thereby enabling the mini-program plug-in to be processed more intelligently, accurately and efficiently.

[0056] In order to process mini-program plug-ins more intelligently, accurately and efficiently, so as to improve the user experience and the convenience of function expansion, the present application provides an embodiment of a mini-program plug-in processing method, see Figure 1 The mini program plug-in processing method specifically includes the following contents:

[0057] Step S101: receiving a mini-program plug-in editing request sent by a user, extracting keywords and functional features from the mini-program plug-in editing request according to the mini-program plug-in editing request and a preset natural language processing algorithm, and determining corresponding plug-in functional summary features according to the keywords and functional features;

[0058] Optionally, in this embodiment, in step S101, the system receives a mini-program plug-in editing request sent by a user, and extracts keywords and functional features from the request and a preset natural language processing algorithm, and finally determines the corresponding plug-in functional summary features. The following is the specific implementation process, technical principles, problems solved, and technical effects of this step.

[0059] First, after the user sends a mini-program plug-in editing request, the system needs to parse the request. Mini-program plug-in editing requests usually contain descriptions of the functions that the user wants to achieve. These descriptions may involve the specific purpose of the plug-in, target users, expected effects, etc. In order to effectively extract this information, the system needs to use natural language processing (NLP) algorithms to conduct in-depth analysis of the request text.

[0060] In terms of technical implementation, the system first receives the editing request text sent by the user and preprocesses it. This includes removing irrelevant characters, standardizing the text format, etc. Next, the system uses a preset natural language processing algorithm to segment the text and divide the text into several words or phrases. Word segmentation is a basic step in natural language processing. By splitting continuous text into independent words, subsequent analysis and processing can be better performed.

[0061] Next, the system needs to extract keywords and functional features from the word segmentation results. Keywords are words that can reflect the main content of the text, usually with high frequency and importance. Functional features are specific descriptions related to the plug-in function, which may include verbs, noun phrases, etc. In order to accurately extract keywords and functional features, the system can use a variety of technical methods, such as TF-IDF (term frequency-inverse document frequency) algorithm, word vector models (such as Word2Vec, GloVe), etc. These methods can evaluate the possibility of keywords and functional features based on the frequency and importance of words in the text.

[0062] In terms of technical principles, the TF-IDF algorithm determines the importance of a word by calculating the frequency of occurrence of a word in a document and its inverse frequency in the entire document set. The word vector model captures the semantic relationship between words by mapping them into a high-dimensional vector space, thereby extracting functional features. These technical methods can effectively extract keywords and functional features that reflect user needs from text.

[0063] In terms of the technical problems to be solved, step S101 mainly solves the problem of how to accurately extract the plug-in function requirements from the user's natural language description. When describing the plug-in function, users often use natural language, which is a flexible and changeable description method and difficult to directly convert into a format that can be understood and processed by machines. By using natural language processing algorithms, the system can automatically parse the user description, extract keywords and functional features, and thus accurately understand the user's needs.

[0064] In terms of the technical effects achieved, through step S101, the system can automatically and accurately extract the plug-in function requirements from the user's natural language description. This not only improves the system's ability to understand user needs, but also reduces the tedious steps for users to describe plug-in functions, making the user experience more friendly and efficient.

[0065] For example, suppose a user wants to develop a weather forecast plug-in and sends an editing request like this: "I need a plug-in that can display the current weather and the weather forecast for the next 7 days, and it is best to provide a weather warning function." After receiving this request, the system first pre-processes the text, removes irrelevant characters and standardizes the format. Then, the system uses a natural language processing algorithm to segment the text and obtain words such as "I", "need", "one", "can", "display", "current weather", "next 7 days", "weather forecast", "plug-in", "best", "can also", "provide", "weather warning", and "function".

[0066] Then, the system uses the TF-IDF algorithm and word vector model to analyze these words and determine the keywords and functional features. Possible keywords include "display", "current weather", "next 7 days", "weather forecast", "weather warning", etc., while functional features include "display weather", "provide weather warning", etc. Based on these keywords and functional features, the system can determine the plug-in function summary features that the user wants to achieve, namely "display current weather, forecast weather for the next 7 days, and provide weather warning function".

[0067] Through this process, the system can accurately understand the user's needs and provide clear functional profile features for subsequent plug-in development. This not only improves development efficiency, but also ensures that the final developed plug-in can meet the user's needs.

[0068] To give another example, suppose a user wants to develop a plug-in for managing to-do lists and sends an editing request like this: "Please help me design a plug-in that can create, edit, and delete to-do lists, and it would be best if it can also set a reminder function." After receiving this request, the system will also pre-process the text first, and then use the natural language processing algorithm to segment the words to obtain words such as "please", "help me", "design", "one", "can", "create", "edit", "delete", "to-do list", "plug-in", "best", "can also", "set", "reminder", and "function".

[0069] Next, the system uses the TF-IDF algorithm and word vector model to analyze these words and extract keywords and functional features. Possible keywords include "create", "edit", "delete", "to-do", "set reminder", etc., while functional features include "create to-do", "edit to-do", "delete to-do", "set reminder function", etc. Based on these keywords and functional features, the system can determine the plug-in function summary features that the user wants to achieve, namely "create, edit and delete to-do, set reminder function".

[0070] Through this process, the system can accurately understand the user's needs and provide clear functional outline features for subsequent plug-in development. This not only improves development efficiency, but also ensures that the final developed plug-in can meet the user's needs and improve user experience and satisfaction.

[0071] In general, through step S101, the system can use the natural language processing algorithm to extract keywords and functional features from the user's natural language description, accurately understand the user's plug-in function requirements, and determine the corresponding functional summary features. This process not only improves the system's ability to understand user needs, but also simplifies the steps for users to describe plug-in functions, making the user experience more friendly and efficient.

[0072] Step S102: inputting the plug-in function summary feature into the feature selection layer of the preset decision tree model, determining the corresponding decision tree selection feature according to the preset segmentation rule, inputting the decision tree selection feature into the splitting layer of the preset decision tree model for recursive splitting operation, determining the corresponding decision tree recursive feature, and determining the corresponding decision tree prediction feature output by the preset decision tree model according to the decision tree recursive feature;

[0073] Optionally, in this embodiment, step S102 involves inputting the plug-in function summary features into a preset decision tree model, and finally determining the decision tree prediction features output by the decision tree model through processing at the feature selection layer and the splitting layer. The specific implementation process, technical principles, problems solved, and technical effects are as follows.

[0074] First, the system inputs the plug-in function summary features extracted in step S101 into the feature selection layer of the preset decision tree model. The decision tree model is a common machine learning model suitable for classification and regression tasks. The main function of the feature selection layer is to select the most useful features for the decision tree model prediction from the input features according to the preset segmentation rules. The preset segmentation rules can measure the importance of each feature based on statistical measurements (such as information gain, Gini index, etc.) and select the optimal features for node splitting.

[0075] In terms of technical implementation, the system first passes the plug-in function summary features as input to the feature selection layer of the decision tree model. The feature selection layer evaluates each feature according to the preset segmentation rules. For example, information gain measures the improvement in purity brought by a feature when dividing a data set, while the Gini index measures the degree to which a feature reduces the impurity of a data set. By calculating these indicators, the feature selection layer can determine which features contribute most to the prediction and select these features as decision tree selection features.

[0076] Next, the system inputs the decision tree selection features into the splitting layer of the decision tree model for recursive splitting operations. The role of the splitting layer is to split the data set according to the value of the selection feature to form nodes and branches of the tree. Recursive splitting is the core of decision tree construction. By continuously selecting the optimal features for data segmentation, a tree structure is eventually formed, and each leaf node represents a final prediction result.

[0077] In terms of technical principles, the recursive splitting operation is achieved by selecting the best splitting feature at each node and dividing the data set into several subsets. After each split, the system repeats the feature selection and splitting operations on the new subset until the stopping condition is met (such as reaching the maximum tree depth or leaf node purity). This recursive process can effectively decompose complex decision problems into a series of simple decision steps and improve the prediction accuracy of the model.

[0078] In terms of the technical problems to be solved, step S102 mainly solves the problem of how to extract the most useful features for prediction from the plug-in function summary features, and how to build a decision tree model that can make accurate predictions. Through the processing of the feature selection layer and the splitting layer, the system can automatically select the optimal features and build a decision tree structure, thereby improving the prediction ability of the model.

[0079] In terms of the technical effects achieved, through step S102, the system can extract the most useful features from the plug-in function summary features and build an efficient decision tree model. In this way, the system can use the decision tree model to accurately predict the plug-in function in subsequent steps, thereby improving the accuracy of plug-in function recommendations and user satisfaction.

[0080] For example, suppose a user wants to develop a health monitoring plug-in and sends an editing request: "I need a plug-in that can record steps, monitor heart rate, and analyze sleep quality." The plug-in function summary features extracted by the system in step S101 may include "recording steps", "monitoring heart rate", "analyzing sleep quality", etc. Next, in step S102, the system inputs these features into the feature selection layer of the preset decision tree model.

[0081] The feature selection layer will evaluate these features according to the preset segmentation rules (such as information gain) to determine the optimal feature. If the "Monitoring Heart Rate" feature brings the highest information gain when segmenting the data set, the system will select this feature as the decision tree selection feature. Then, the system will input the "Monitoring Heart Rate" feature into the splitting layer of the decision tree model for recursive splitting operations.

[0082] In the split layer, the system divides the dataset into different subsets based on the value of the "monitor heart rate" feature. For example, the system might divide the dataset into two subsets with heart rate above a certain threshold and below the threshold. The system then repeats the feature selection and splitting operations on each subset, selecting other features (such as "log steps" or "analyze sleep quality") for further splitting until the complete decision tree is built.

[0083] Through this process, the system finally determines the corresponding decision tree recursive features, and determines the decision tree prediction features output by the decision tree model based on these features. For example, the system may output a prediction feature indicating that the user needs a health plug-in that focuses on monitoring heart rate.

[0084] For another example, suppose the user wants to develop a financial management plug-in and sends an editing request: "Please help me design a plug-in that can record expenses, generate budgets, and analyze consumption trends." The plug-in function summary features extracted by the system in step S101 may include "record expenses", "generate budgets", "analyze consumption trends", etc. Next, in step S102, the system inputs these features into the feature selection layer of the preset decision tree model.

[0085] The feature selection layer evaluates these features based on preset segmentation rules (such as the Gini index) to determine the optimal feature. If the "Generate Budget" feature brings the largest reduction in the Gini index when segmenting the data set, the system will select this feature as the decision tree selection feature. The system then inputs the "Generate Budget" feature into the splitting layer of the decision tree model for a recursive splitting operation.

[0086] In the splitting layer, the system divides the dataset into different subsets based on the value of the "generate budget" feature. For example, the system might divide the dataset into two subsets with budgets above a certain threshold and below the threshold. The system then repeats the feature selection and splitting operations on each subset, selecting other features (such as "record spending" or "analyze consumption trends") for further splitting until a complete decision tree is built.

[0087] Through this process, the system finally determines the corresponding decision tree recursive features, and determines the decision tree prediction features output by the decision tree model based on these features. For example, the system may output a prediction feature indicating that the user needs a financial management plug-in that focuses on generating budgets.

[0088] In general, through step S102, the system can extract the most useful features from the plug-in function summary features, build an efficient decision tree model, and output the decision tree prediction features. This process not only improves the prediction ability of the system, but also ensures that the plug-in finally developed can more accurately meet user needs.

[0089] Step S103: Input the decision tree prediction feature into the feature integration layer of the preset regression tree model, determine the corresponding plug-in classification 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, determine the corresponding plug-in recommendation model according to the plug-in classification feature, the preset real feature and the back propagation algorithm, and push the recommended plug-in output by the plug-in recommendation model to the user.

[0090] Optionally, in this embodiment, step S103 involves inputting the decision tree prediction features into a preset regression tree model, and finally determining the plug-in classification features and plug-in recommendation model through the processing of the feature integration layer and the pruning layer, combined with the preset model integration algorithm, and pushing the recommended plug-in to the user. The following is the specific implementation process, technical principles, problems solved, and technical effects of this step.

[0091] First, the system receives the decision tree prediction features generated in step S102 and inputs them into the feature integration layer of the preset regression tree model. The regression tree model is a common machine learning model suitable for the prediction of continuous values. The function of the feature integration layer is to integrate the decision tree prediction features with other related features (such as user historical data, plug-in usage frequency, etc.) to generate a comprehensive feature vector. The integrated feature vector can more comprehensively reflect user needs and plug-in functional characteristics, and improve the prediction ability of the model.

[0092] In terms of technical implementation, the system first inputs the decision tree prediction features and other related features into the feature integration layer. The feature integration layer combines these features through methods such as feature splicing and weighted averaging to generate a comprehensive feature vector. Then, the system inputs the comprehensive feature vector into the pruning layer of the regression tree model. The main function of the pruning layer is to prune the regression tree model through preset pruning rules (such as the minimum description length principle, cross-validation, etc.), remove overfitting branches, and retain the optimal regression tree structure.

[0093] In terms of technical principles, the regression tree model continuously selects the optimal segmentation point to segment data and gradually builds a tree structure. The pruning layer uses a pruning algorithm to evaluate the contribution of each branch to the model prediction and removes branches that contribute less to the prediction, thereby simplifying the model structure and improving the generalization ability of the model. In this way, the system can obtain a simplified and efficient regression tree model.

[0094] Next, the system determines the corresponding plug-in classification features based on the regression tree prediction features output by the pruning layer and the preset model integration algorithm. The model integration algorithm can use a variety of methods, such as weighted average, voting method, stacking method, etc., to integrate the prediction results of different models to generate the final plug-in classification features. The plug-in classification features are used to represent the category or type of plug-in, such as tools, entertainment, health, etc. By integrating the prediction results of multiple models, the accuracy and robustness of classification can be improved.

[0095] In terms of the technical problems to be solved, step S103 mainly solves the problem of how to generate accurate plug-in classification features from the comprehensive feature vector, and how to build a plug-in recommendation model that can make accurate recommendations. Through the processing of the feature integration layer and the pruning layer, the system can automatically generate comprehensive feature vectors and regression tree prediction features, and generate accurate plug-in classification features through the model integration algorithm.

[0096] In terms of the technical effects achieved, through step S103, the system can generate a comprehensive feature vector from the decision tree prediction features and other related features, build an efficient regression tree model, and output accurate plug-in classification features. In this way, the system can use the plug-in classification features and preset real features in subsequent steps to build an accurate plug-in recommendation model, thereby improving the accuracy of plug-in recommendations and user satisfaction.

[0097] For example, suppose a user wants to develop a health monitoring plug-in and sends an editing request: "I need a plug-in that can record steps, monitor heart rate, and analyze sleep quality." The decision tree prediction features generated by the system in step S102 may include "step recording", "heart rate monitoring", "sleep analysis", etc. Next, in step S103, the system inputs these prediction features into the feature integration layer of the preset regression tree model.

[0098] The feature integration layer integrates these prediction features with the user's historical data (such as previously used health plug-ins, usage frequency, etc.) to generate a comprehensive feature vector. The system then inputs the comprehensive feature vector into the pruning layer of the regression tree model, and after pruning, a simplified regression tree prediction feature is obtained. Next, the system integrates the regression tree prediction features through a preset model integration algorithm (such as the weighted average method) to generate the final plug-in classification feature, such as "health monitoring category."

[0099] Based on the plug-in classification features and preset real features (such as the functions and user reviews of existing health monitoring plug-ins on the market), the system uses the back propagation algorithm to build a plug-in recommendation model. The back propagation algorithm optimizes the recommendation model by continuously adjusting the model parameters to minimize the error between the predicted results and the actual results. Finally, the system pushes the recommended plug-in (such as a well-known health monitoring plug-in) output by the plug-in recommendation model to the user.

[0100] As another example, suppose the user wants to develop a financial management plug-in and sends an editing request: "Please help me design a plug-in that can record expenditures, generate budgets, and analyze consumption trends." The decision tree prediction features generated by the system in step S102 may include "expenditure records", "budget generation", "consumption analysis", etc. Next, in step S103, the system inputs these prediction features into the feature integration layer of the preset regression tree model.

[0101] The feature integration layer integrates these prediction features with the user's historical data (such as previously used financial plug-ins, usage frequency, etc.) to generate a comprehensive feature vector. The system then inputs the comprehensive feature vector into the pruning layer of the regression tree model, and after pruning, a simplified regression tree prediction feature is obtained. Next, the system integrates the regression tree prediction features through a preset model integration algorithm (such as voting method) to generate the final plug-in classification feature, such as "Financial Management".

[0102] Based on the plug-in classification features and preset real features (such as the functions and user reviews of existing financial management plug-ins on the market), the system uses the back propagation algorithm to build a plug-in recommendation model. The back propagation algorithm optimizes the recommendation model by continuously adjusting the model parameters to minimize the error between the predicted results and the actual results. Finally, the system pushes the recommended plug-in (such as a well-known financial management plug-in) output by the plug-in recommendation model to the user.

[0103] In general, through step S103, the system can generate plug-in classification features from the comprehensive feature vector, build an accurate plug-in recommendation model, and push the recommended plug-ins to the user.

[0104] From the above description, it can be seen that the mini-program plug-in processing method provided in the embodiment of the present application can receive the mini-program plug-in editing request sent by the user, extract keywords and functional features from the mini-program plug-in editing request according to the mini-program plug-in editing request and the preset natural language processing algorithm, and determine the corresponding plug-in function summary features according to the keywords and functional features; input the decision tree prediction features into the feature integration layer of the preset regression tree model, determine the corresponding plug-in classification 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, determine the corresponding plug-in recommendation model according to the plug-in classification features, the preset real features and the back propagation algorithm, and push the recommended plug-in output by the plug-in recommendation model to the user, thereby being able to process the mini-program plug-in more intelligently, accurately and efficiently.

[0105] In one embodiment of the applet plug-in processing method of the present application, see Figure 2 , and can also include the following:

[0106] Step S201: performing a minority class identification operation on the plug-in function summary feature according to a preset minority class identification rule, and determining a corresponding nearest neighbor function feature according to the minority class function feature obtained after the minority class identification operation and a preset K nearest neighbor algorithm;

[0107] Step S202: determining a corresponding balanced functional feature according to the minority class functional feature and the nearest neighbor functional feature.

[0108] Optionally, in this embodiment, the execution order of steps S201 and S202 is key, and they together constitute the process of identifying and optimizing plug-in functional features. First, in step S201, the plug-in functional summary features are subjected to minority class recognition operations through preset minority class recognition rules. The purpose of minority class recognition is to identify features that account for a small proportion but are important in the functional summary features. Minority class recognition rules can be based on statistical methods or machine learning algorithms, such as SMOTE (Synthetic Minority Over-sampling Technique) or other balancing methods. Through minority class recognition operations, the system can extract minority class functional features from the functional summary features, which appear less frequently in the overall data set but may be of great significance.

[0109] In the technical process, the system first loads the functional summary features of the plug-in, and then applies the preset minority class identification rules. The minority class identification rules analyze the distribution of functional features and mark and extract those features with low frequency but important. Next, the system determines the corresponding nearest neighbor functional features based on the minority class functional features and the preset K nearest neighbor algorithm. The K nearest neighbor algorithm is a common classification and regression algorithm that finds the K neighbors closest to the target data point by calculating the distance between data points (such as Euclidean distance, Manhattan distance, etc.). By comparing the features of these neighbors, the system can infer the category or feature value of the target data point.

[0110] In terms of technical principles, the K nearest neighbor algorithm uses the principle of "birds of a feather flock together", that is, similar data points tend to have similar features. The minority class identification rule identifies those features that appear less frequently in the overall data set but are of great significance through analysis of data distribution. Combining the two, the system can effectively identify minority class functional features in the plug-in function summary features and find their corresponding nearest neighbor functional features through the K nearest neighbor algorithm. This process not only improves the accuracy of feature recognition, but also ensures the representativeness and diversity of the features.

[0111] In terms of the technical problems to be solved, step S201 mainly solves the problem of how to identify minority functional features from a large number of functional summary features. Minority functional features often appear less frequently in the overall data set, but they may have an important impact on the function and performance of the plug-in. Through the minority class identification operation, the system can effectively extract these important minority functional features and determine their corresponding nearest neighbor functional features through the K nearest neighbor algorithm, thereby improving the accuracy and representativeness of feature recognition.

[0112] In terms of the technical effects achieved, through step S201, the system can identify minority class function features from the plug-in function summary features, and find its nearest neighbor function features through the K nearest neighbor algorithm. This process not only improves the accuracy of feature recognition, but also ensures the representativeness and diversity of the features, thus laying the foundation for subsequent feature optimization and model training.

[0113] Next, the system enters step S202, and determines the corresponding balanced functional features according to the minority class functional features and the nearest neighbor functional features. The determination of balanced functional features is to avoid model bias caused by uneven feature distribution in the subsequent model training and prediction process. The system generates a balanced functional feature set by combining the minority class functional features and the nearest neighbor functional features. These feature sets include both minority class functional features and similar functional features, thereby improving the representativeness and diversity of the features.

[0114] In the technical process, the system first conducts a comprehensive analysis of the minority class features and the nearest neighbor features. Through balancing algorithms (such as weighted average, undersampling, oversampling, etc.), the system balances the minority class features and the nearest neighbor features to generate a new feature set. The choice of balancing algorithm can be adjusted according to the specific application scenario and data distribution to ensure that the generated balanced features are both representative and can effectively avoid model bias.

[0115] In terms of technical principles, the determination of balanced functional features utilizes the similarity between minority class functional features and nearest neighbor functional features, and these features are processed comprehensively through the balanced algorithm to generate a representative and diverse functional feature set. The balanced algorithm samples and weights the data to ensure that the generated functional feature set has a uniform distribution in the overall data set, thereby improving the generalization ability and prediction accuracy of the model.

[0116] In terms of the technical problems to be solved, step S202 mainly solves the problem of how to balance the minority class features and the nearest neighbor features. Through the balancing algorithm, the system can effectively avoid the model deviation caused by the uneven distribution of features and improve the generalization ability and prediction accuracy of the model.

[0117] In terms of the technical effects achieved, through step S202, the system can generate a balanced set of functional features, which includes both minority functional features and similar functional features, thereby improving the representativeness and diversity of the features. By determining the balanced functional features, the system can avoid model deviations caused by uneven feature distribution in the subsequent model training and prediction process, and improve the generalization ability and prediction accuracy of the model.

[0118] For example, suppose a user wants to develop an educational plug-in and sends an editing request: "Please help me design a plug-in that can identify students' learning behaviors and analyze learning effects." The minority class functional features generated by the system in step S201 may include "learning behavior identification", "learning effect analysis", etc. These features rarely appear in the overall data set, but have an important impact on the function and performance of the plug-in. Next, the system uses the preset K nearest neighbor algorithm to find the nearest neighbor functional features of these minority class functional features, such as "behavior analysis", "effect evaluation", etc.

[0119] In step S202, the system balances the minority class features and the nearest neighbor features to generate a balanced feature set. These feature sets contain both minority class features and similar features, thereby improving the representativeness and diversity of the features. By determining the balanced features, the system can avoid model deviations caused by uneven feature distribution in subsequent model training and prediction processes, and improve the generalization ability and prediction accuracy of the model. Finally, based on the balanced features, the system constructs an educational plug-in that can accurately identify students' learning behaviors and analyze learning effects, thereby improving user satisfaction and experience.

[0120] In one embodiment of the applet plug-in processing method of the present application, see Figure 3 , and can also include the following:

[0121] Step S301: performing a distance calculation operation on the minority class functional feature and the nearest neighbor functional feature to determine the corresponding functional feature distance;

[0122] Step S302: determining a corresponding synthetic functional feature according to the functional feature distance and a preset ratio rule, performing a feature adding operation on the functional feature according to the synthetic functional feature, and determining a corresponding balancing functional feature.

[0123] Optionally, in this embodiment, in step S301, the system first performs a distance calculation operation on the minority class functional features and the nearest neighbor functional features to determine the functional feature distance between them. The distance calculation operation is based on the similarity measurement of feature vectors, and their similarity is quantified by calculating the distance between two feature vectors. Common distance calculation methods include Euclidean distance, Manhattan distance, cosine similarity, etc. The system first represents the minority class functional features and the nearest neighbor functional features as feature vectors, and then selects a suitable distance calculation method to calculate the distance between them. The purpose of this process is to quantify the degree of difference between the minority class functional features and the nearest neighbor functional features, thereby providing a basis for subsequent feature synthesis and optimization.

[0124] In the technical process, the system first converts the minority class function features and the nearest neighbor function features into feature vector representations. These feature vectors can be numerical representations of the plug-in functions, such as function frequency, number of uses, etc. Next, the system selects a suitable distance calculation method, such as Euclidean distance, and calculates the distance between the feature vectors. The result of the distance calculation will be used to measure the similarity between the minority class function features and the nearest neighbor function features. In this way, the system is able to quantify the differences between the function features and provide a basis for subsequent feature synthesis.

[0125] In terms of technical principles, the distance calculation operation utilizes the spatial relationship between feature vectors. By calculating the distance between feature vectors, the system can quantify the similarity between them. Euclidean distance is a common distance calculation method that quantifies similarity by calculating the straight-line distance between feature vectors. Other methods such as cosine similarity measure similarity by calculating the angle between vectors. Choosing a suitable distance calculation method can more accurately reflect the similarity between functional features, thereby improving the effect of feature synthesis.

[0126] In terms of the technical problems solved, step S301 mainly solves the problem of how to quantify the similarity between the minority class function features and the nearest neighbor function features. Through the distance calculation operation, the system can accurately quantify the differences between the function features, providing a basis for subsequent feature synthesis and optimization. This process ensures the accuracy of feature synthesis and avoids the failure of feature synthesis due to inaccurate feature similarity.

[0127] In terms of the technical effect achieved, through step S301, the system can accurately quantify the similarity between the minority class function features and the nearest neighbor function features, providing a basis for subsequent feature synthesis. The results of the distance calculation operation will be used to guide feature synthesis and optimization, ensuring that the generated synthetic function features have high representativeness and accuracy.

[0128] Next, the system enters step S302, determines the corresponding synthetic functional features according to the functional feature distance and the preset ratio rule, and performs feature addition operation on the functional features according to the synthetic functional features, so as to determine the corresponding balanced functional features. The synthetic functional features are determined by combining the minority class functional features and the nearest neighbor functional features to generate a new functional feature. This process can adopt a variety of methods, such as weighted average, linear interpolation, etc. The preset ratio rule is used to guide the generation of the synthetic functional features to ensure that the synthesis results have high representativeness and accuracy.

[0129] In the technical process, the system first determines the weights of minority class functional features and nearest neighbor functional features based on the functional feature distance and preset ratio rules. Then, the system synthesizes these features according to the preset ratio to generate new synthetic functional features. The method for generating synthetic functional features can be selected according to the specific application scenario. For example, the weighted average method can generate synthetic functional features based on the inverse of the functional feature distance as the weight. The generated synthetic functional features will be used for subsequent feature addition operations.

[0130] In terms of technical principles, the generation of synthetic functional features utilizes the similarity between minority class functional features and nearest neighbor functional features. By presetting proportional rules, the system can reasonably assign feature weights and generate representative synthetic functional features. Weighted averaging is a common synthesis method that uses the inverse of feature distance as the weight to perform weighted averaging of minority class functional features and nearest neighbor functional features to generate synthetic functional features. Other methods such as linear interpolation generate new feature vectors through linear combinations.

[0131] In terms of the technical problems to be solved, step S302 mainly solves the problem of how to generate representative synthetic functional features and use them for feature addition operations. Through the generation of synthetic functional features, the system can effectively integrate minority class functional features and nearest neighbor functional features to improve the representativeness and accuracy of the features. The feature addition operation generates balanced functional features by adding synthetic functional features to the original functional feature set, ensuring the balance and diversity of the feature set.

[0132] In terms of the technical effects achieved, through step S302, the system can generate representative synthetic functional features and use them for feature addition operations to generate balanced functional features. This process not only improves the representativeness and accuracy of the functional features, but also ensures the balance and diversity of the feature set, thereby providing a reliable feature basis for subsequent model training and prediction.

[0133] For example, suppose a user wants to develop a smart home control plug-in and sends an editing request: "Please help me design a plug-in that can control lights, adjust temperature, and monitor security." The minority class functional features generated by the system in step S301 may include "lighting control", "temperature adjustment", "security monitoring", etc. These features rarely appear in the overall data set, but have an important impact on the function and performance of the plug-in. Then, the system calculates the distance between the minority class functional features and the nearest neighbor functional features, such as "lighting control", "environmental adjustment", etc. through distance calculation operations.

[0134] In step S302, the system generates synthetic functional features, such as "intelligent lighting control" and "comfortable temperature adjustment", based on the functional feature distance and preset proportion rules. Then, the system adds these synthetic functional features to the original functional feature set to generate balanced functional features. Through this process, the system can generate a representative and accurate functional feature set, providing a reliable feature basis for subsequent model training and prediction. Finally, based on the balanced functional features, the system builds a smart home control plug-in that can accurately control lighting, adjust temperature, and monitor safety, thereby improving user satisfaction and experience.

[0135] In one embodiment of the applet plug-in processing method of the present application, see Figure 4 , and can also include the following:

[0136] Step S401: inputting the balancing function feature into the feature selection layer of a preset decision tree model, and determining the corresponding feature segmentation point according to a preset Gini impurity;

[0137] Step S402: determining corresponding decision tree selection features according to the feature segmentation points and the balance function features.

[0138] Optionally, in this embodiment, in step S401, the system inputs the balancing function feature into the feature selection layer of the preset decision tree model, and determines the corresponding feature segmentation point according to the preset Gini impurity. The feature selection layer is an important component of the decision tree model, and its function is to select the best feature for segmentation by calculating the Gini impurity of each feature. Gini impurity is an indicator to measure the purity of a data set, and the lower the value, the purer the data set. Specifically, the system first inputs the balancing function feature into the decision tree model, and then calculates the Gini impurity of each feature. The calculation method of Gini impurity is to traverse the possible segmentation points of all features, calculate the Gini impurity of each segmentation point, and select the segmentation point that can minimize the Gini impurity as the feature segmentation point.

[0139] In the technical process, the system first loads the balancing function features, which are generated by the previous steps and have high representativeness and balance. Then, the system inputs these features into the feature selection layer of the decision tree model. The feature selection layer traverses all possible split points. For each split point, it calculates the Gini impurity of the left and right child nodes, and combines the Gini impurity of the split point to obtain the total Gini impurity of the current split point. The system selects the split point with the lowest Gini impurity as the best feature split point. In this way, the system can determine the best feature split point, thus providing a basis for the construction of the decision tree model.

[0140] In terms of technical principles, the feature selection layer uses the calculation method of Gini impurity. By traversing all possible segmentation points of the features, it selects the segmentation point that can minimize the Gini impurity. Gini impurity is an indicator used to measure the purity of a data set. The lower its value, the purer the samples in the data set. By selecting the segmentation point with the lowest Gini impurity, the system can maximize the classification effect of the decision tree model. This process of the feature selection layer ensures that the decision tree model can select the best feature segmentation point at each step, thereby improving the accuracy and generalization ability of the model.

[0141] In terms of the technical problems solved, step S401 mainly solves the problem of how to select the best feature segmentation point in the balanced functional features. By calculating the Gini impurity, the system can measure the classification effect of each feature and select the segmentation point that can minimize the Gini impurity. This process ensures that the decision tree model can select the best feature for segmentation at each step, thereby improving the classification effect and accuracy of the model.

[0142] In terms of the technical effects achieved, through step S401, the system can select the best feature segmentation point from the balanced functional features to improve the classification effect and accuracy of the decision tree model. The feature selection layer calculates the Gini impurity to ensure that the feature segmentation points selected in each step are optimal, thereby improving the accuracy and generalization ability of the model. This process not only improves the performance of the model, but also effectively avoids overfitting.

[0143] Next, the system enters step S402 to determine the corresponding decision tree selection features based on the feature segmentation points and the balance function features. Decision tree selection features means that at each node, the system selects features for segmentation based on the feature segmentation points determined in the previous step. Specifically, the system first obtains the feature segmentation points determined in step S401, and then selects corresponding features from the balance function features based on these segmentation points. The process of selecting features is to traverse the balance function features and select features that can maximize the classification effect based on the segmentation points of each feature.

[0144] In the technical process, the system first obtains the feature segmentation points determined in step S401. These segmentation points are determined by calculating the Gini impurity and have a higher classification effect. Then, the system traverses the balanced functional features and selects the features that can maximize the classification effect according to the segmentation points of each feature. The process of selecting features is to select the features with the best classification effect by comparing the segmentation effect of each feature. In this way, the system can determine the selected features of the decision tree model at each node, thereby improving the classification effect and accuracy of the model.

[0145] In terms of technical principles, the decision tree feature selection process utilizes the calculation method of feature segmentation points and classification effects. By traversing the balanced functional features, the system can compare the segmentation effect of each feature and select the feature with the best classification effect. The feature selection process ensures that the decision tree model can select the best feature for segmentation at each node, thereby improving the classification effect and accuracy of the model. This process not only improves the performance of the model, but also effectively avoids overfitting.

[0146] In terms of the technical problems solved, step S402 mainly solves the problem of how to select the best features for segmentation in the decision tree model. By comparing the segmentation effect of each feature, the system can select the feature with the best classification effect, thereby improving the classification effect and accuracy of the model. This process ensures that the decision tree model can select the best feature for segmentation at each node, thereby improving the performance and accuracy of the model.

[0147] In terms of the technical effects achieved, through step S402, the system can select the best features for segmentation in the decision tree model, thereby improving the classification effect and accuracy of the model. The process of selecting features ensures that the decision tree model can select the best features for segmentation at each node by comparing the segmentation effect of each feature, thereby improving the performance and accuracy of the model. This process not only improves the classification effect of the model, but also effectively avoids overfitting, thereby improving the generalization ability of the model.

[0148] For example, suppose a user wants to develop a health monitoring plug-in and sends an editing request: "Please help me design a plug-in that can monitor heart rate, blood pressure and sleep quality." The balance function features generated by the system in step S401 may include "heart rate monitoring", "blood pressure monitoring", "sleep quality analysis", etc. Then, the system determines the best segmentation point for each feature by calculating the Gini impurity, such as the segmentation point of heart rate may be "70 beats per minute", and the segmentation point of blood pressure may be "blood pressure 120 / 80", etc.

[0149] In step S402, the system selects corresponding features from the balance function features based on these segmentation points, such as selecting the "heart rate monitoring" feature based on "heart rate 70 beats per minute" and selecting the "blood pressure monitoring" feature based on "blood pressure 120 / 80". In this way, the system can determine the selected features of the decision tree model at each node, thereby improving the classification effect and accuracy of the model. Finally, based on these selected features, the system builds a health monitoring plug-in that can accurately monitor heart rate, blood pressure, and sleep quality, thereby improving user satisfaction and experience.

[0150] In one embodiment of the applet plug-in processing method of the present application, see Figure 5 , and can also include the following:

[0151] 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;

[0152] 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.

[0153] Optionally, in this embodiment, in step S501, the system inputs the decision tree selection feature into the splitting layer of the preset decision tree model, and performs a recursive splitting operation according to the preset maximum tree depth, thereby determining the corresponding decision tree recursive feature. The splitting layer is an important component of the decision tree model, and its main function is to divide the data set into purer subsets through recursive splitting operations. The recursive splitting operation refers to starting from the root node and continuously selecting the best features for splitting until the preset maximum tree depth or other stopping conditions are reached. The preset maximum tree depth is to prevent overfitting and ensure the generalization ability of the model.

[0154] The specific technical process is that the system first loads the decision tree selection features determined in the previous step. These features have been screened by the feature selection layer and are highly representative. Then, the system inputs these features into the splitting layer of the decision tree model. The splitting layer starts from the root node and performs recursive splitting operations layer by layer according to the preset maximum tree depth. At each layer, the splitting layer selects the best feature for splitting, generates left and right child nodes, and continues to split each child node until the preset maximum tree depth or other stopping conditions are reached. In this process, the system continuously calculates the purity of each node, selects the features that can maximize the purity for splitting, and thus determines the corresponding decision tree recursive features.

[0155] In terms of technical principles, the split layer uses recursive splitting operations and purity calculation methods to recursively divide the data set into purer subsets layer by layer by selecting the best features for splitting. The recursive splitting operation is the core step of the decision tree model. By continuously selecting the best features for splitting, the system can maximize the purity of the nodes, thereby improving the classification effect of the model. The preset maximum tree depth is to prevent the model from being too complex, avoid overfitting, and ensure that the model has good generalization ability.

[0156] In terms of the technical problems solved, step S501 mainly solves the problem of how to perform recursive splitting operations in the decision tree model. Through the recursive splitting operation, the system can divide the data set layer by layer into purer subsets, thereby improving the classification effect and accuracy of the model. The preset maximum tree depth ensures that the model is not too complex, avoids overfitting problems, and makes the model have better generalization ability.

[0157] In terms of the technical effects achieved, through step S501, the system can determine the corresponding decision tree recursive features from the decision tree selection features, thereby improving the classification effect and accuracy of the decision tree model. The recursive split operation selects the best features for splitting, ensuring that each step of the split is optimal, thereby improving the accuracy and generalization ability of the model. The preset maximum tree depth prevents the model from being too complex and ensures the generalization ability and stability of the model.

[0158] Next, the system enters step S502 to perform a pruning operation on the decision tree recursive features to determine the decision tree prediction features output by the corresponding preset decision tree model. The pruning operation refers to simplifying the decision tree model by deleting some unnecessary nodes or branches to improve the generalization ability and stability of the model. The decision tree prediction features refer to the features that are ultimately used to predict the category label. These features are retained after the pruning operation and have a higher classification effect and representativeness.

[0159] The specific technical process is that the system first obtains the recursive features of the decision tree determined in the previous step. These features have a higher classification effect after recursive splitting operations. Then, the system performs pruning operations on these features. The pruning process is to traverse all nodes of the decision tree, calculate the importance of each node, and delete those nodes or branches that contribute less to the model. The system determines which nodes or branches can be deleted by calculating the importance of the nodes, thereby simplifying the model structure and improving the generalization ability and stability of the model. Finally, the system retains those important nodes and branches and determines the corresponding decision tree prediction features.

[0160] In terms of technical principles, the pruning operation utilizes the node importance calculation method to simplify the decision tree model structure by deleting nodes or branches that contribute less to the model. Node importance is measured by calculating the contribution of each node to the model classification effect. The higher the contribution, the higher the importance of the node. Through the pruning operation, the system can delete those unimportant nodes or branches, thereby simplifying the model structure and improving the generalization ability and stability of the model.

[0161] In terms of the technical problems to be solved, step S502 mainly solves the problem of how to prune the decision tree model. Through the pruning operation, the system can delete unnecessary nodes or branches, simplify the model structure, and improve the generalization ability and stability of the model. The pruning operation can also improve the computational efficiency of the model, reduce the risk of overfitting, and make the model more stable in practical applications.

[0162] In terms of the technical effects achieved, through step S502, the system can perform pruning operations on the decision tree recursive features, determine the corresponding decision tree prediction features, and improve the generalization ability and stability of the model. The pruning operation simplifies the model structure by deleting unnecessary nodes or branches, and improves the computational efficiency and stability of the model. The final decision tree prediction features have high classification effect and representativeness, and can be effectively used to predict category labels.

[0163] For example, suppose a user wants to develop a plug-in for identifying plant species and sends an editing request: "Please help me design a plug-in that can identify different plant species." The decision tree selection features generated by the system in step S501 may include "leaf shape", "flower color", "leaf vein structure", etc. Then, the system determines the decision tree recursive features through recursive splitting operations according to the preset maximum tree depth, such as splitting different child nodes according to "leaf shape" and further splitting purer subsets according to "flower color".

[0164] In step S502, the system prunes these recursive features, deletes nodes or branches that contribute less to the classification effect, and simplifies the model structure. The decision tree prediction features that are finally retained may be "leaf shape" and "flower color", which have high classification effect and representativeness. In this way, the system can build an efficient and stable plant identification plug-in that can accurately identify different types of plants and improve user satisfaction and experience.

[0165] In one embodiment of the applet plug-in processing method of the present application, see Figure 6 , and can also include the following:

[0166] Step S601: inputting the balance function feature into the feature selection layer of the preset regression tree model, performing feature selection operation according to the preset mean square error algorithm, determining the corresponding regression tree selection feature, performing recursive splitting operation according to the regression tree selection feature, and determining the corresponding regression tree recursive feature;

[0167] 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.

[0168] Optionally, in this embodiment, in step S601, the system inputs the balance function feature into the feature selection layer of the preset regression tree model, and performs feature selection operation according to the preset mean square error algorithm to determine the corresponding regression tree selection feature. The regression tree model is a machine learning algorithm for predicting continuous values, and its core lies in minimizing the prediction error by recursively splitting the data set. In the feature selection layer, the system uses the mean square error algorithm to evaluate the prediction effect of each feature, so as to select those features that contribute most to the model prediction effect.

[0169] The specific technical process is that the system first receives the input balance function features, which have been preprocessed to remove noise and outliers. Then, the system inputs these features into the feature selection layer of the regression tree model. The feature selection layer uses the mean square error algorithm to evaluate each feature. The mean square error algorithm measures the prediction error by calculating the square difference between the predicted value and the actual value. The smaller the error, the better the prediction effect of the feature. The system selects those features with the smallest error as the regression tree selection features based on the mean square error value of the feature.

[0170] In terms of technical principles, the mean square error algorithm is a commonly used regression model evaluation method that can effectively measure the prediction effect of features. The feature selection layer calculates the mean square error value of each feature and selects those features with the smallest error, thereby improving the prediction accuracy of the model. The regression tree model divides the data set into purer subsets layer by layer through recursive splitting operations, thereby minimizing the prediction error and improving the prediction effect of the model.

[0171] In terms of the technical problems solved, step S601 mainly solves the problem of how to perform feature selection and recursive splitting operations in the regression tree model. Through the mean square error algorithm, the system can effectively evaluate the prediction effect of each feature and select those features that contribute the most to the model, thereby improving the prediction accuracy of the model. The recursive splitting operation can further optimize the model structure, divide the data set layer by layer into purer subsets, and minimize the prediction error.

[0172] In terms of the technical effects achieved, through step S601, the system can determine the corresponding regression tree selection features from the balance function features, and determine the regression tree recursive features through recursive splitting operations to improve the prediction accuracy of the model. The mean square error algorithm ensures the accuracy of feature selection, and the recursive splitting operation optimizes the model structure, enabling it to better adapt to the data and improve the prediction effect.

[0173] Next, the system enters step S602, inputs the regression tree recursive features into the pruning layer of the preset regression tree model, performs pruning operations on these features, and determines the corresponding regression tree prediction features. Pruning operation refers to simplifying the regression tree model by deleting some unnecessary nodes or branches to improve the generalization ability and stability of the model. The pruned regression tree prediction features will be input into the feature integration layer for the final continuous numerical prediction.

[0174] The specific technical process is that the system first obtains the recursive features of the regression tree determined in the previous step. These features have a high prediction effect after the recursive splitting operation. Then, the system inputs these features into the pruning layer of the regression tree model. The pruning layer calculates the importance of each node and deletes those nodes or branches that contribute less to the model, thereby simplifying the model structure. The system traverses all nodes of the regression tree, calculates the importance of each node, and determines which nodes or branches can be deleted, thereby improving the generalization ability and stability of the model. Finally, the system retains those important nodes and branches and determines the corresponding regression tree prediction features.

[0175] In terms of technical principles, the pruning operation uses the node importance calculation method to simplify the regression tree model structure by deleting nodes or branches that contribute less to the model. Node importance is measured by calculating the contribution of each node to the model prediction effect. The higher the contribution, the higher the importance of the node. Through the pruning operation, the system can delete those unimportant nodes or branches, thereby simplifying the model structure and improving the generalization ability and stability of the model.

[0176] In terms of the technical problems to be solved, step S602 mainly solves the problem of how to prune the regression tree model. Through the pruning operation, the system can delete unnecessary nodes or branches, simplify the model structure, and improve the generalization ability and stability of the model. The pruning operation can also improve the computational efficiency of the model, reduce the risk of overfitting, and make the model more stable in practical applications.

[0177] In terms of the technical effects achieved, through step S602, the system can prune the regression tree recursive features, determine the corresponding regression tree prediction features, and input them into the feature integration layer to improve the generalization ability and stability of the model. The pruning operation simplifies the model structure and improves the computational efficiency and stability of the model by deleting unnecessary nodes or branches. The regression tree prediction features finally determined have high prediction effect and representativeness, and can be effectively used to predict continuous values.

[0178] For example, suppose a user wants to develop a plug-in for predicting housing prices and sends an editing request: "Please help me design a plug-in that can predict housing prices." The regression tree selection features generated by the system in step S601 may include "house area", "house age", "geographic location", etc. Then, the system determines the regression tree recursive features through recursive splitting operations and according to the preset mean square error algorithm, such as splitting different child nodes according to "house area" and further splitting a purer subset according to "geographic location".

[0179] In step S602, the system prunes these recursive features, deletes nodes or branches that contribute less to the prediction effect, and simplifies the model structure. The regression tree prediction features that are finally retained may be "house area" and "geographic location", which have high prediction effect and representativeness. In this way, the system can build an efficient and stable house price prediction plug-in that can accurately predict house prices and improve user satisfaction and experience.

[0180] In one embodiment of the applet plug-in processing method of the present application, see Figure 7 , and can also include the following:

[0181] 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;

[0182] 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 plug-in classification feature.

[0183] Optionally, in this embodiment, in step S701, the system needs to 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. The core of this step is to reasonably allocate weights by analyzing the features output by two different models so that each feature can play its maximum role in the final prediction process.

[0184] The specific technical process is that the system first obtains the decision tree prediction features and regression tree prediction features that have been determined in the previous steps. These features have undergone feature selection, recursive splitting and pruning in their respective models and have a high prediction effect. Next, the system needs to perform weight matching operations on these features. The process of weight matching operation is to calculate the importance of each feature in different models and assign corresponding weights. Importance can be measured by indicators such as the contribution of the feature in the model, the frequency of occurrence, and the prediction error. The system calculates the weight of each feature based on these indicators to ensure that the decision tree prediction features and the regression tree prediction features can balance their contributions in the final feature weighted averaging process.

[0185] In terms of technical principles, the weight matching operation utilizes the feature importance calculation method, which analyzes the performance of each feature in the model and reasonably allocates weights. The importance of a feature is determined by calculating its contribution to the prediction result. The higher the contribution, the greater the importance of the feature. In this way, the system can ensure that the weight of each feature is reasonably allocated and give full play to its role in the prediction process.

[0186] In terms of the technical problems to be solved, step S701 mainly solves the problem of how to match weights between the feature outputs of the decision tree model and the regression tree model. Due to the different working principles and feature selection processes of the decision tree model and the regression tree model, directly weighting the features of the two may lead to poor prediction results. Through the weight matching operation, the system can reasonably allocate the weight of each feature, ensuring that the final feature weighted average result can balance the advantages of the two models and improve the prediction accuracy.

[0187] In terms of the technical effects achieved, through step S701, the system can perform weight matching operations based on the decision tree prediction features and the regression tree prediction features to determine the corresponding decision tree prediction feature weights and regression tree prediction feature weights. Reasonable weight distribution ensures the contribution of each feature in the final prediction process and improves the accuracy and stability of the prediction model. Ultimately, the system can fully utilize the advantages of the two models in the feature weighted averaging process to achieve better prediction results.

[0188] Next, the system enters step S702, and performs a feature weighted average operation based on the decision tree prediction feature weight and the regression tree prediction feature weight to determine the corresponding plug-in classification feature. The feature weighted average operation refers to generating a comprehensive feature set for the final prediction by performing a weighted average on the decision tree prediction feature and the regression tree prediction feature. This comprehensive feature set can better reflect the characteristics of the data and improve the prediction accuracy and stability of the model.

[0189] The specific technical process is that the system first obtains the decision tree prediction feature weights and regression tree prediction feature weights determined in the previous step. Then, the system performs a weighted average operation on these features. The weighted average operation process is to multiply each feature by its corresponding weight, and then sum them to generate a comprehensive feature value. In this way, the system fuses the decision tree prediction features and the regression tree prediction features to generate a comprehensive plug-in classification feature.

[0190] In terms of technical principles, the feature weighted average operation utilizes the weighted average algorithm to generate a comprehensive feature value by weighting each feature. The weighted average algorithm can effectively integrate the features of different models, ensure that the weight distribution of each feature is reasonable, and improve the prediction effect of the comprehensive feature set. In this way, the system can generate a comprehensive feature set with high prediction accuracy and stability.

[0191] In terms of the technical problems solved, step S702 mainly solves the problem of how to perform weighted averaging of the prediction features of the decision tree and the prediction features of the regression tree. Since the feature weights of the two models are different, direct feature fusion may lead to poor prediction results. Through the weighted average operation, the system can reasonably allocate the weight of each feature, generate a comprehensive feature set, and improve the prediction accuracy and stability of the model.

[0192] In terms of the technical effects achieved, through step S702, the system can perform a feature weighted average operation based on the decision tree prediction feature weights and the regression tree prediction feature weights to determine the corresponding plug-in classification features. The weighted average operation ensures that the weight distribution of each feature is reasonable, and the generated comprehensive feature set can better reflect the characteristics of the data and improve the prediction accuracy and stability of the model. Finally, the system can generate a plug-in classification feature with a higher prediction effect for the final prediction.

[0193] For example, suppose a user wants to develop a plug-in for classifying customer credit risk and sends an editing request: "Please help me design a plug-in that can classify customer credit risk." The decision tree prediction features generated by the system in step S701 may include "income level", "credit history", "debt ratio", etc., and the regression tree prediction features may include "age", "loan amount", "repayment record", etc. Then, the system assigns corresponding weights according to the importance of each feature through weight matching operations, such as "income level" has a higher weight in the decision tree model, and "repayment record" has a higher weight in the regression tree model.

[0194] In step S702, the system performs a weighted average operation on the decision tree prediction features and the regression tree prediction features according to these weights to generate a comprehensive plug-in classification feature. In this way, the system can build an efficient and stable customer credit risk classification plug-in that can accurately classify customer credit risks and improve user satisfaction and experience.

[0195] In order to process mini-program plug-ins more intelligently, accurately and efficiently, so as to improve user experience and the convenience of function expansion, the present application provides an embodiment of a mini-program plug-in processing device for implementing all or part of the contents of the mini-program plug-in processing method, see Figure 8 The mini-program plug-in processing device specifically includes the following contents:

[0196] The plug-in collection module 10 is used to collect the historical mini-program access data sent by the user, perform feature extraction operations on the historical mini-program access data according to preset feature extraction rules, determine the corresponding plug-in function summary features, and determine the corresponding balance function features according to the plug-in function summary features and a preset feature balance algorithm;

[0197] The plug-in classification module 20 is used to input the balance function feature into the feature selection layer of the preset decision tree model, determine the corresponding decision tree selection feature according to the preset segmentation rule, input the decision tree selection feature into the 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;

[0198] The plug-in recommendation module 30 is used to input the decision tree prediction features into the feature integration layer of the preset regression tree model, determine the corresponding plug-in classification 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, determine the corresponding mini-program push model according to the plug-in classification features, the preset real features and the back propagation algorithm, and perform the mini-program push operation according to the mini-program push model.

[0199] From the above description, it can be seen that the mini-program plug-in processing device provided in the embodiment of the present application can receive the mini-program plug-in editing request sent by the user, extract keywords and functional features from the mini-program plug-in editing request according to the mini-program plug-in editing request and the preset natural language processing algorithm, and determine the corresponding plug-in function summary features according to the keywords and functional features; input the decision tree prediction feature into the feature integration layer of the preset regression tree model, determine the corresponding plug-in classification 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, determine the corresponding plug-in recommendation model according to the plug-in classification feature, the preset real feature and the back propagation algorithm, and push the recommended plug-in output by the plug-in recommendation model to the user, thereby being able to process the mini-program plug-in more intelligently, accurately and efficiently.

[0200] From the hardware level, in order to process mini-program plug-ins more intelligently, accurately and efficiently to improve user experience and the convenience of function expansion, the present application provides an embodiment of an electronic device for implementing all or part of the contents of the mini-program plug-in processing method, and the electronic device specifically includes the following contents:

[0201] 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 information transmission between the applet plug-in processing device and related devices such as the core business system, user terminal and related database; the logic controller can be a desktop computer, a tablet computer and a mobile terminal, etc., but the present embodiment is not limited thereto. In the present embodiment, the logic controller can be implemented with reference to the embodiment of the applet plug-in processing method and the embodiment of the applet plug-in processing device in the embodiment, and the contents thereof are incorporated herein, and the repeated parts are not repeated.

[0202] 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.

[0203] In practical applications, part of the applet plug-in processing method 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.

[0204] 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.

[0205] 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.

[0206] In one embodiment, the mini-program plug-in processing method function may be integrated into the central processing unit 9100. The central processing unit 9100 may be configured to perform the following control:

[0207] Step S101: receiving a mini-program plug-in editing request sent by a user, extracting keywords and functional features from the mini-program plug-in editing request according to the mini-program plug-in editing request and a preset natural language processing algorithm, and determining corresponding plug-in functional summary features according to the keywords and functional features;

[0208] Step S102: inputting the plug-in function summary feature into the feature selection layer of the preset decision tree model, determining the corresponding decision tree selection feature according to the preset segmentation rule, inputting the decision tree selection feature into the splitting layer of the preset decision tree model for recursive splitting operation, determining the corresponding decision tree recursive feature, and determining the corresponding decision tree prediction feature output by the preset decision tree model according to the decision tree recursive feature;

[0209] Step S103: Input the decision tree prediction feature into the feature integration layer of the preset regression tree model, determine the corresponding plug-in classification 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, determine the corresponding plug-in recommendation model according to the plug-in classification feature, the preset real feature and the back propagation algorithm, and push the recommended plug-in output by the plug-in recommendation model to the user.

[0210] From the above description, it can be seen that the electronic device provided in the embodiment of the present application receives a mini-program plug-in editing request sent by a user, extracts keywords and functional features from the mini-program plug-in editing request according to the mini-program plug-in editing request and a preset natural language processing algorithm, and determines the corresponding plug-in function summary features according to the keywords and functional features; inputs the decision tree prediction features into the feature integration layer of the preset regression tree model, determines the corresponding plug-in classification 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, determines the corresponding plug-in recommendation model according to the plug-in classification features, the preset real features, and the back propagation algorithm, and pushes the recommended plug-in output by the plug-in recommendation model to the user, thereby being able to process the mini-program plug-in more intelligently, accurately, and efficiently.

[0211] In another embodiment, the mini-program plug-in processing device can be configured separately from the central processing unit 9100. For example, the mini-program plug-in processing device can be configured as a chip connected to the central processing unit 9100, and the mini-program plug-in processing method function is implemented through the control of the central processing unit.

[0212] like Fig. 9As 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.

[0213] like Fig. 9 As 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.

[0214] 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.

[0215] 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.

[0216] 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.

[0217] 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.).

[0218] 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.

[0219] 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. The communication module 9110 (transmitter / receiver) 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.

[0220] The embodiments of the present application also provide a computer-readable storage medium capable of implementing all steps of the applet plug-in processing method in the above embodiments, 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 steps of the applet plug-in processing method in the above embodiments are implemented. For example, when the processor executes the computer program, the following steps are implemented:

[0221] Step S101: receiving a mini-program plug-in editing request sent by a user, extracting keywords and functional features from the mini-program plug-in editing request according to the mini-program plug-in editing request and a preset natural language processing algorithm, and determining corresponding plug-in functional summary features according to the keywords and functional features;

[0222] Step S102: inputting the plug-in function summary feature into the feature selection layer of the preset decision tree model, determining the corresponding decision tree selection feature according to the preset segmentation rule, inputting the decision tree selection feature into the splitting layer of the preset decision tree model for recursive splitting operation, determining the corresponding decision tree recursive feature, and determining the corresponding decision tree prediction feature output by the preset decision tree model according to the decision tree recursive feature;

[0223] Step S103: Input the decision tree prediction feature into the feature integration layer of the preset regression tree model, determine the corresponding plug-in classification 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, determine the corresponding plug-in recommendation model according to the plug-in classification feature, the preset real feature and the back propagation algorithm, and push the recommended plug-in output by the plug-in recommendation model to the user.

[0224] From the above description, it can be seen that the computer-readable storage medium provided in the embodiment of the present application receives a mini-program plug-in editing request sent by a user, extracts keywords and functional features from the mini-program plug-in editing request according to the mini-program plug-in editing request and a preset natural language processing algorithm, and determines the corresponding plug-in function summary features according to the keywords and functional features; inputs the decision tree prediction features into the feature integration layer of the preset regression tree model, determines the corresponding plug-in classification 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, determines the corresponding plug-in recommendation model according to the plug-in classification features, the preset real features, and the back propagation algorithm, and pushes the recommended plug-in output by the plug-in recommendation model to the user, thereby enabling the mini-program plug-in to be processed more intelligently, accurately, and efficiently.

[0225] The embodiments of the present application also provide a computer program product capable of implementing all steps of the applet plug-in processing method in the above embodiments, where the execution subject is a server or a client. When the computer program / instruction is executed by a processor, the steps of the applet plug-in processing method are implemented. For example, the computer program / instruction implements the following steps:

[0226] Step S101: receiving a mini-program plug-in editing request sent by a user, extracting keywords and functional features from the mini-program plug-in editing request according to the mini-program plug-in editing request and a preset natural language processing algorithm, and determining corresponding plug-in functional summary features according to the keywords and functional features;

[0227] Step S102: inputting the plug-in function summary feature into the feature selection layer of the preset decision tree model, determining the corresponding decision tree selection feature according to the preset segmentation rule, inputting the decision tree selection feature into the splitting layer of the preset decision tree model for recursive splitting operation, determining the corresponding decision tree recursive feature, and determining the corresponding decision tree prediction feature output by the preset decision tree model according to the decision tree recursive feature;

[0228] Step S103: Input the decision tree prediction feature into the feature integration layer of the preset regression tree model, determine the corresponding plug-in classification 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, determine the corresponding plug-in recommendation model according to the plug-in classification feature, the preset real feature and the back propagation algorithm, and push the recommended plug-in output by the plug-in recommendation model to the user.

[0229] From the above description, it can be seen that the computer program product provided by the embodiment of the present application receives a mini-program plug-in editing request sent by a user, extracts keywords and functional features from the mini-program plug-in editing request according to the mini-program plug-in editing request and a preset natural language processing algorithm, and determines the corresponding plug-in function summary features according to the keywords and functional features; inputs the decision tree prediction features into the feature integration layer of the preset regression tree model, determines the corresponding plug-in classification 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, determines the corresponding plug-in recommendation model according to the plug-in classification features, the preset real features, and the back propagation algorithm, and pushes the recommended plug-in output by the plug-in recommendation model to the user, thereby being able to process the mini-program plug-in more intelligently, accurately, and efficiently.

[0230] It should be understood 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 aspects. 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.

[0231] 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.

[0232] 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 including 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.

[0233] 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 in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0234] 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 method for processing a mini-program plug-in, characterized in that: The method comprises: Receive a mini-program plug-in editing request sent by a user, extract keywords and functional features from the mini-program plug-in editing request according to the mini-program plug-in editing request and a preset natural language processing algorithm, and determine corresponding plug-in functional summary features according to the keywords and functional features; Input the plug-in function summary feature into the feature selection layer of the preset decision tree model, determine the corresponding decision tree selection feature according to the preset segmentation rule, input the decision tree selection feature into the 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; The decision tree prediction feature is input into the feature integration layer of the preset regression tree model, and the corresponding plug-in classification feature is determined 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. The corresponding plug-in recommendation model is determined according to the plug-in classification feature, the preset real feature, and the back propagation algorithm, and the recommended plug-in output by the plug-in recommendation model is pushed to the user.

2. The applet plug-in processing method according to claim 1, characterized in that: 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: 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; 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.

3. The applet plug-in processing method according to claim 1, characterized in that: The determining of the corresponding plug-in classification 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 average operation is performed according to the decision tree prediction feature weight and the regression tree prediction feature weight to determine the corresponding plug-in classification feature.

4. A small program plug-in processing device, characterized in that: The device comprises: A plug-in collection module, used to receive a mini-program plug-in editing request sent by a user, extract keywords and functional features from the mini-program plug-in editing request according to the mini-program plug-in editing request and a preset natural language processing algorithm, and determine corresponding plug-in functional summary features according to the keywords and functional features; A plug-in classification module, used to input the plug-in function summary 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; A plug-in recommendation module is used to input the decision tree prediction features into the feature integration layer of the preset regression tree model, determine the corresponding plug-in classification 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, determine the corresponding plug-in recommendation model according to the plug-in classification features, the preset real features, and the back propagation algorithm, and push the recommended plug-in output by the plug-in recommendation model to the user.

Images