State Analysis Of Time Series Data Based On Fluctuation Vector And Constryction Of Expert System Model

As the Internet wave promotes the rapid development of related industries toward industrial intelligence and information digitization,various industries have generated and accumulated massive amounts of data.The amount of data has shown an exponential upward trend,and the information age has entered the era of big data.As an important branch of big data research,time series data analysis can obtain key information in the operation of the system,and has broad application prospects in various industries.It has received a lot of attention in the research of related fields at home and abroad.The mutation point detection technology can quickly locate the mutation point from the overall data,and analyze the cause of the data fluctuation based on the position information of the mutation point.Because the traditional data stream mutation point detection technology has problems such as high time-consuming and weak anti-interference ability when processing large-scale time series data,it cannot effectively meet the needs of data stream detection.In order to effectively improve the rapid analysis of the disease state of time series data,this paper provides a rapid detection model for multiple mutation points based on the combination of the TSTKS mutation point detection algorithm and the sliding window;the template matching and membership analysis method of the fluctuation vector is adopted.,Proposed a system architecture for rapid analysis of time series data and status diagnosis;based on time series data under epilepsy and other pathological conditions,an expert system model for pathological data status analysis was initially constructed to realize data abnormal status detection,epileptic pathological status classification,and pathological changes.Diagnosis of abnormal detection.Firstly,based on the combination of the TSTKS mutation point detection algorithm and the sliding window method,a rapid detection model for multiple mutation points is given.The volatility of each window is extracted as the window distribution feature through the multi-mutation point detection model of the TTSKS algorithm,and multiple continuous volatility is defined as a volatility vector.A template matching and membership analysis algorithm based on the volatility vector is proposed for rapid detection and analysis of abnormal time series data.analysis.Secondly,using the template matching and membership analysis method of fluctuation vector,a system architecture for rapid analysis and status diagnosis of time series data is proposed.According to the different fluctuation characteristics of epilepsy EEG signals in the onset state,the actual epilepsy data is selected for experiments,and the method of template matching and membership analysis of fluctuation vectors is used to establish the corresponding standard template library for different disease states of epilepsy EEG signals.Rapid detection of lesion data,through comparative analysis with template matching methods based on correlation coefficient and Manhattan distance.Finally,based on the time series data under brain epilepsy and other pathological conditions,an expert system model for analyzing the status of pathological data is preliminarily constructed.Based on the template matching and membership analysis method based on the fluctuation vector and the multi-classification study based on the disease state,an expert system model based on the analysis of the disease state is constructed,and the knowledge base and the reasoning mechanism of the epilepsy disease diagnosis expert system model are designed.The rapid detection of largescale lesion data and the classification of local lesion states are used to assist pathological analysis and diagnosis.The relationship between the optimal sliding window width and the dimension of the fluctuation vector is determined through the simulation experiment data,and the feasibility of the fluctuation vector as a rapid detection of the state of the time series data is verified.The experimental results show that the template matching method proposed in this paper has more advantages and better detection effect.In addition,for the three main epileptic lesions of patients with epilepsy,the support vector machine multi-classifier is introduced to extract the features of the lesions to achieve the classification of the lesions.Experiments show that this method has a significant effect on the state classification of epilepsy lesion data,can quickly locate the lesion data and make accurate reasoning diagnosis of the lesion state in a short time,and can be used to assist physicians in pathological analysis and diagnosis.