Application system fault prediction method based on association rule and deep learning integrated model

Abstract

An application system fault prediction method based on an association rule and a deep learning integrated model, the method comprising: collecting historical fault data of an application system; performing data preprocessing on the collected historical fault data to obtain a data matrix, and then converting the data matrix into a data form suitable for analysis by an association rule algorithm and a deep learning algorithm; performing feature extraction on the data matrix to obtain effective fault-related features; establishing an association rule model and a deep learning LSTM model; fusing outputs of the association rule model and the deep learning LSTM model to form a multi-layer decision tree model; and predicting a fault by combining the association rule model, the deep learning LSTM model, the multi-layer decision tree model, and current state data of the application system, and obtaining a fault prediction result via weighted voting. The present invention achieves accurate fault prediction and early warning for an application system, and enhances the accuracy and robustness of fault prediction by introducing optimizations for a deep learning LSTM model.

Description

An application system fault prediction method based on association rules and deep learning integrated model

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to the Chinese patent application filed with the State Intellectual Property Office of China on December 29, 2023, with application number 202311854819X and application name “A method for predicting application system faults based on an integrated model of association rules and deep learning”, the entire contents of which are incorporated by reference into this application. Technical Field

[0003] The present invention relates to the field of deep learning LSTM models, and in particular to an application system fault prediction method based on an integrated model of association rules and deep learning. Background Art

[0004] As application systems continue to grow in complexity and criticality, the impact of failures on system operation becomes increasingly significant. Traditional fault prediction methods often rely on statistical and rule-based analysis, but they often fail to accurately predict failures, especially when dealing with complex fault scenarios. Summary of the Invention

[0005] The purpose of this invention is to provide an application system fault prediction method based on an integrated model of association rules and deep learning. By utilizing historical fault data and real-time status data, by mining association rules and establishing a deep learning model integration, accurate prediction and early warning of application system faults can be achieved. By introducing the optimization of the deep learning LSTM model, the accuracy and robustness of fault prediction are enhanced.

[0006] A method for predicting application system failures based on an integrated model of association rules and deep learning, comprising the following steps:

[0007] Collect historical fault data of the application system, including fault occurrence time, fault type, fault cause, fault duration, and fault impact range;

[0008] The collected historical fault data is preprocessed to obtain a data matrix, which is then converted into a data form suitable for analysis by association rule algorithms and deep learning algorithms;

[0009] Perform feature extraction on the data matrix to obtain effective features related to the fault;

[0010] Based on the association rule algorithm, frequent item sets and association rules are mined from historical fault data. By establishing strong association rules, potential fault causes and patterns are revealed, and an association rule model is obtained. Based on the deep learning algorithm and the extracted effective fault-related features, the temporal relationships and temporal patterns in the fault data are captured, and a deep learning LSTM model is established.

[0011] Based on the association rule model and the deep learning LSTM model, a multi-layer decision tree ensemble learning method is used to fuse the outputs of the association rule model and the deep learning LSTM model to form a multi-layer decision tree model;

[0012] The association rule model, deep learning LSTM model, multi-layer decision tree model and the status data of the current application system are combined to predict faults. The prediction results of the association rule model, deep learning LSTM model and multi-layer decision tree model are integrated through weighted voting to obtain the fault prediction results.

[0013] Furthermore, the conversion into a data form suitable for analysis by an association rule algorithm and a deep learning LSTM model specifically includes: the association rule algorithm converts the data matrix into an association rule transaction set, where each row in the association rule transaction set represents a transaction and each column represents an item; the deep learning LSTM model converts the data matrix into LSTM sequence data, where each row in the LSTM sequence data represents a sequence, each column represents a time step, and each time step contains multiple features.

[0014] Furthermore, the effective features related to the fault include statistical features of the fault occurrence time and coding features of the fault type.

[0015] Furthermore, the deep learning LSTM model adopts a long short-term memory network structure to capture the temporal relationships and temporal patterns in fault data.

[0016] Furthermore, when training the deep learning LSTM model, hyperparameter tuning and supervised learning algorithms are used to improve the accuracy and robustness of the predictions.

[0017] Furthermore, when monitoring the application system status data in real time, a data stream joining method is adopted to update the multi-layer decision tree model in real time to maintain the prediction accuracy.

[0018] Furthermore, it also includes: taking emergency measures in a timely manner according to the fault prediction results, including fault handling, maintenance and backup, to reduce the impact of the fault on the application system.

[0019] Furthermore, the application system includes a server, network equipment, and a database system.

[0020] The present invention collects historical fault data from application systems, including information such as fault occurrence time, fault type, and fault cause. Then, through data preprocessing and feature extraction, the raw data is converted into a form suitable for analysis by a deep learning LSTM model. Next, fault prediction training and modeling are performed using an association rule model and an optimized deep learning LSTM model. While monitoring the application system's status data in real time, data streams are added to promptly update the prediction model to maintain high prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] FIG1 is a flowchart of an application system fault prediction method based on an integrated model of association rules and deep learning according to an embodiment of the present invention;

[0022] FIG2 is a diagram showing the architecture and data processing flow of an embodiment of the present invention. DETAILED DESCRIPTION

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0024] As shown in FIG1 and FIG2 , an embodiment of the present invention provides a fault prediction method based on an integrated model of association rules and deep learning, comprising the following steps:

[0025] Step 101: Collect historical fault data, which includes fault occurrence time, fault type, fault cause, fault duration, fault impact range, etc.

[0026] Step 102: Perform data preprocessing and feature extraction on the collected historical fault data, converting it into a format suitable for analysis using the association rule algorithm and deep learning LSTM model. Specifically, after data preprocessing, a data matrix is ​​generated, consisting of multiple features such as fault occurrence time, fault type, and fault cause. The association rule algorithm converts the data matrix into an association rule transaction set, where each row represents a transaction and each column represents an item. The deep learning LSTM model converts the data matrix into sequence data, where each row represents a sequence and each column represents a time step, with each time step containing multiple features.

[0027] The preprocessing includes data cleaning, denoising, and outlier processing to improve the quality and usability of the data. The data matrix after preprocessing is shown in Table 1.

[0028] Table 1

[0029] The association rule transaction set after data matrix conversion is shown in Table 2.

[0030] Table 2

[0031] The data matrix is ​​converted into LSTM sequence data as shown in Table 3.

[0032] Table 3

[0033] Step 103: Extract features from the data matrix to obtain effective features related to the fault. These features can be statistical features of the fault occurrence time, coding features of the fault type, etc.

[0034] Step 104: Use the association rule model for mining. Based on the Apriori algorithm, unrelated single transaction items are deleted to reduce the number of scans of the transaction database, improve the efficiency and accuracy of the algorithm, and mine frequent item sets and association rules in the fault data. By establishing strong association rules, potential fault causes and patterns are revealed for fault prediction and early warning.

[0035] Step 105: Based on the association rule model, a deep learning LSTM model is introduced for optimization. The statistical features of fault occurrence time and the encoded features of fault type extracted during the data processing phase are used as the input layer of the deep neural network and fed into the model along with other features. Through the long short-term memory (LSTM) network structure and multiple layers of nonlinear transformations, the temporal relationships and patterns in the fault data are captured. When training the LSTM model, appropriate hyperparameter tuning and supervised learning algorithms are used to improve prediction accuracy and robustness.

[0036] Step 106: Use the prediction results of the association rule model and the LSTM model as input to train a multi-layer decision tree model.

[0037] Step 107: Monitor the status data of the application system in real time, and use the association rule model, LSTM model, and multi-layer decision tree model to predict faults. The prediction results of the three models are fused using a weighted voting method. The final fusion model is as follows:

[0038] Among them, y is the final fault prediction result, n is the number of base models, and w i is the weight of the i-th base model, h i (x) is the prediction result of the i-th base model for the input x, and sign is the sign function used to convert the prediction result into a binary classification label.

[0039] Step 108: Collect equipment operation data (fault occurrence time, fault duration, fault type, fault cause, etc.) in real time and compare and match it with the model prediction data to verify the accuracy of the prediction, and optimize and adjust the model parameters based on the verification results.

[0040] Step 109: Based on the prediction results, take corresponding emergency measures, including fault handling, maintenance, and backup, to promptly reduce the impact of the fault on the application system.

[0041] Through the above implementation, the present invention provides an efficient and accurate method for predicting application system failures based on an association rule model. This method leverages the advantages of the association rule model and the deep learning LSTM model, achieving better results in handling complex application system failures.

[0042] The present invention uses a confusion matrix to evaluate the accuracy of the integrated model's predictions. A confusion matrix is ​​a table that displays the correspondence between the model's predictions and the actual results. Each row of the confusion matrix represents the actual fault type, and each column represents the fault type predicted by the model. The elements on the diagonal of the confusion matrix represent the number of times the model's predictions were correct, while the elements on the off-diagonal represent the number of times the model's predictions were incorrect.

[0043] The model is validated using 100 application faults as a test dataset. If there are four fault types, the confusion matrix may be as shown in Table 4.

[0044] Table 4

[0045] Accuracy = (25 + 19 + 23 + 24) / (100) = 0.91

[0046] Precision = 25 / (25+2+1+2) = 0.83

[0047] Recall = 25 / (25+1+1+2) = 0.86

[0048] F1 value = 2*0.83*0.86 / (0.83+0.86) = 0.84

[0049] In the case of 100 test samples, the model accuracy can reach more than 90%, and the model performance index can reach more than 80%, with good generalization ability and prediction effect.

[0050] Compared with the traditional method, the present invention has the following advantages:

[0051] 1. Improved accuracy: By mining frequent itemsets and association rules in historical fault data and combining it with a deep learning LSTM model, we can more comprehensively analyze the rules and patterns in fault data and improve the accuracy of fault prediction.

[0052] 2. Enhanced robustness: The optimized deep learning LSTM model can effectively capture the temporal relationships and patterns in fault data, enabling the prediction results to better adapt to different types of fault conditions and improving the robustness of the prediction results.

[0053] 3. Real-time optimization: By adding data streams, the prediction model is updated in real time, which can promptly adapt to changes in application system status data, maintain prediction accuracy, and quickly respond to potential failure risks.

[0054] 4. Broad Applicability: The method of the present invention is applicable to various application systems, including servers, network equipment, database systems, etc. The method can be trained and modeled based on historical fault data of different systems, thereby realizing personalized fault prediction solutions.

[0055] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. An application system fault prediction method based on an integrated model of association rules and deep learning, characterized in that Including the following steps: Collect historical fault data of the application system, where the historical fault data includes the fault occurrence time, fault type, fault cause, fault duration, and fault impact scope; Perform data preprocessing on the collected historical fault data to obtain a data matrix, and then convert it into a data form suitable for analysis by association rule algorithms and deep learning algorithms; Extract features from the data matrix to obtain effective features related to faults; Based on the association rule algorithm, mine frequent item sets and association rules from historical fault data, and reveal potential fault causes and patterns by establishing strong association rules to obtain an association rule model; based on the deep learning algorithm and the effective features related to faults extracted, capture the temporal relationship and temporal pattern in the fault data, and establish a deep learning LSTM model; Based on the association rule model and the deep learning LSTM model, use the multi-layer decision tree ensemble learning method to fuse the outputs of the association rule model and the deep learning LSTM model to form a multi-layer decision tree model; Combine the association rule model, the deep learning LSTM model, the multi-layer decision tree model and the status data of the current application system to predict faults, and fuse the prediction results of the association rule model, the deep learning LSTM model and the multi-layer decision tree model through weighted voting to obtain the fault prediction result.

2. The method for predicting application system faults based on an association rule model according to claim 1, wherein The conversion into a data form suitable for analysis by the association rule algorithm and the deep learning LSTM model specifically includes: the association rule algorithm converts the data matrix into an association rule transaction set, where each row in the association rule transaction set represents a transaction and each column represents an item; the deep learning LSTM model converts the data matrix into LSTM sequence data, where each row in the LSTM sequence data represents a sequence, each column represents a time step, and each time step contains multiple features.

3. The method for predicting application system failures based on an association rule model according to claim 1, characterized in that The effective features related to faults include the statistical features of the fault occurrence time and the encoded features of the fault type.

4. The method for predicting application system faults based on an association rule model according to claim 1, wherein, The deep learning LSTM model adopts a long short-term memory network structure to capture the temporal relationship and temporal pattern in the fault data.

5. The method for predicting application system faults based on an association rule model according to claim 1, wherein When training the deep learning LSTM model, use hyperparameter tuning and supervised learning algorithms to improve the accuracy and robustness of prediction.

6. The method for predicting application system faults based on an association rule model according to claim 1, wherein When real-time monitoring the status data of the application system, adopt the method of data stream addition to update the multi-layer decision tree model in real time to maintain the prediction accuracy.

7. The method for predicting application system faults based on the association rule model according to claim 1, wherein, It also includes: Take emergency measures in a timely manner according to the fault prediction result, including fault handling, maintenance and backup, to reduce the impact of the fault on the application system.

8. The method for predicting application system faults based on an association rule model according to claim 1, wherein, The application system includes servers, network devices, and database systems.

Images