Research On Time Series Stream Event Acquisition And Classification

During the continuous progress of science and technology,the explosive growth of data information involves all fields of human beings.As one of the time series data types,time series stream data is a dynamic data set whose size increases with time.The information carried by the time series stream data can be transmitted in real time,and each data point arrives in time order.The sequence of data points is independent of system scheduling and control and is only related to the sampling time.Because of the fluidity of its data,not only the data scale is huge,but also the data peak is unpredictable.Traditional data analysis and for classification and time sequence flow events on the acquisition and classification approach of different,the data processing is a traditional to randomly without considering the order of the data analysis method,and the time sequence data flow is as the growth of the time constant collection storage,it is a state of flow,each data point with time attributes so the need according to the time label order to deal with the analysis.Through the study of time series flow data,medical health detection,financial bond market analysis,social behavior analysis of users,weather,mine disaster warning,etc.The complete event acquisition and classification of the time series flow events in the time series flow data is the basis to study the operation of this kind of data.At present,there are a lot of researches on the time series flow events.In the traditional data system,the abnormal events are only aimed at the query of the abnormal data points,but in many practical applications,the abnormal events are the exceptions in a period of continuous time,that is,the collection of points in a period of continuous time.It includes a variety of information about time and energy.The traditional threshold method cannot find the abnormal event quickly and obtain the event information completely.There are also some limitations and shortcomings in some existing methods for obtaining abnormal events.To solve this problem,this paper proposes a variable multi-stage time window algorithm to obtain streaming data events.This method firstly USES the STA/LTA algorithm-based event trigger algorithm to find the approximate range of the starting and ending points of the abnormal events,and then USES the AIC rule-based accurate acquisition method to accurately locate the starting and ending points in the estimated range,so as to obtain the complete information of the abnormal events.After the complete abnormal events are obtained,a time series stream event classification method based on neural network is proposed to classify the events.Firstly,the vector data of abnormal events are converted into a matrix of the same size as the specified input form.Meanwhile,the time window size and step size of the proposed method of event sequence stream acquisition are adjusted by feedback after classification,so as to ensure the accuracy of the acquisition and classification criteria.The experimental results show that the proposed method of time series stream event acquisition and classification and some existing algorithms have great advantages in execution efficiency and accuracy.