| In the past years, with the fruitful application of complex network in various fields, the bridges from time series to complex network has defined lots of rules to conserve the information of time series into complex networks, and analyzing some indicators in complex network could reveal some characteristics of time series, which provides a novel view for time series analysis. We retrospect some representative transformations, cycle network, recurrence network, visibility graph, frequency-degree algorithm and random walk network, and introduce their applications. According to this, we have proposed a new method and apply it to ECG analysis.Based on the summary of previous algorithms, we have proposed Auto-Regressive method to transform time series episodes into nodes of complex network, defined the transmitting coefficients between nodes and use adaptive triangular rule to connect the edges. The edges between nodes in complex network accords with time direction and reveal the transmitting patterns.Directed subgraphs in networks constructed by Auto-Regressive methods have reflected transmitting patterns of time series; meanwhile, we introduce the background of the detection and prediction of ECG time series. The three-node directed subgraphs have been applied into detecting and predicting Ventricular Fibrillation (VF) from Normal Sinus Rhythm (NSR) and also compared with other methods. Next, we introduce some about Atrial Fibrillation (AF) and epicardium ECG data; Auto-Regressive method could also be applied into Atrial Fibrillation analysis.Amplitude-temporal method has transformed one data point into one node in complex network, calculated the ration between amplitude difference and time difference, and set a threshold to determine the edges between nodes. Only by average degree, we have applied this method into effectively detecting three kinds of ECG time series, without using the complex indicator of average path length compared with frequency-degree algorithm. |
PDF Full Text Request