Research On Nonlinear Time Series Analysis Method Based On Transition Network

The combination of complex network and nonlinear time series provides a new direction for the nonlinear time series.In recent years,time series complex network construction algorithms have been widely applied to the analysis of nonlinear real data in life such as Internet,biologic information network,traffic flow,weather forecast and so on.In recent years,the application of nonlinear time series analysis method to EEG signal analysis has been favored by some scholars.However,the classification performance of traditional analysis methods for epilepsy EEG is relatively poor,which hinders the application of epileptic classification algorithm in clinical practice.In the first time,the transition network is applied to the analysis of epileptic EEG,and the automatic detection and classification of epileptic EEG is realized.In this paper,we study the construction algorithm of complex network based on time series,and mainly study the algorithm of time series transition network construction.The traditional transition network of time series construction algorithm is summarized,and the algorithm is improved aiming at the shortcomings of traditional algorithm.The epileptic EEG signal is constructed as a transfer network and the classification features are extracted based on the statistical characteristics of the transition network topology,so the high precision automatic detection and classification of epileptic brain signal is realized.Firstly,according to the time series maximum value,the node set of the time series transition network is constructed.Further more,the number of state transfer between nodes is calculated,so the weight matrix is constructed.Then the edge set of the network is determined.The transition network of time series is constructed through the above process.Secondly,transition network of the time series is analyzed by the statistical characteristics of the topology structure of the transition network,and the different types' classification characteristics of time series are proposed in combination with the characteristics of the transition network.The degree change rate,loop coefficient,clustering coefficient and average path length are proposed in this paper.The four characteristics of the extracted time series are analyzed.Finally,the epileptic EEG data is constructed as an improved transition network.Basedon the statistical characteristics of the transition network,four classification features are extracted respectively for the single feature classification of the EEG signals in the epileptic patients' intermittent and episodes of epileptic EEG.The automatic classification of epileptic EEG is realized,and four characteristics are made.The classification performance was evaluated.By comparing with other related experimental results,it has a great improvement on the discrimination accuracy of the two types of epileptic EEG compared with other single feature classification methods in this paper,up to 98.5%.In this paper,by using transition network construction of time series algorithm,epileptic signal is converted into a transition network.And the classification features of excellent performance are extracted to realize high performance epileptic EEG classification.The classification algorithm of Epileptic Electroencephalogram automatic detection proposed in this paper is helpful to promote the research and application of clinical medicine.