Time Series Classification Based on Complex Network

Time series classification plays an important role in various tasks such as electroencephalogram (EEG) classification, electrocardiogram (ECG) classification, human activity recognition and radio signal modulation identification. At present, some promising methods have been proposed to transform time series classification into graph classification by mapping time series to graphs. However, these transformation methods with fixed mapping rules may lack of flexibility, leading to the loss of information, so as to decrease the classification accuracy. In this chapter, we introduce circular limited penetrable visibility graph (CLPVG), a new method of mapping time series to graphs. Furthermore, in order to map time series to graphs more flexibly through deep learning, we also introduce an automatic visibility graph (AVG) based on graph neural network (GNN), a framework which can transform time series into graphs and realize classification end-to-end. Finally, we carry out experiments on some common datasets to prove the effectiveness of our methods.