Real-Time Monitoring And Analysis Of CPU Usage Time Series Data

With the development of economy and technology,monitoring system has gradually changed towards automation and intelligence in the operation scenario.CPU usage time series data is a basic performance indicator for real-time monitoring,and its research on outlier detection is of great significance.However,in the actual scenario,the CPU usage data has the characteristics of strong timeliness,large amount of data,and unknown distribution.The traditional operation methods cannot effectively process and analyze such data.Therefore,the method used in this paper is to extract the key feature attributes based on the sliding window,and then research and analyze the outliers of CPU usage data from the perspective of machine learning.First of all,in the face of the high-frequency data of the CPU usage time series,and considering the limited memory resources of the computer,the data collection method based on the sliding window was adopted,and then data pre-processing was carried out.Secondly,in the face of the time subsequence collected based on the sliding window,it may contain some irrelevant and redundant information.The method of feature extraction was used to reduce the dimension of the subsequence data,and the key information of the subsequence was extracted as the feature attribute of the current time point,including statistical features,fitting features and classification features.Then,for the problem of outlier detection of CPU usage time series data,decision tree algorithm and random forest algorithm were used to classify the feature attribute set separately,that is,whether it is normal or abnormal,and confusion matrix and ROC curve were given for the test results to evaluate the overall effect of the two algorithms.Finally,the experimental results show that the two algorithms perform well on the detection of abnormal points,especially random forest,showing good scalability and robustness,which not only increases the accuracy of the judgment results,but also greatly increases the speed of calculation.