High Dimensional Time Series Classification Based on Multi-Layer Perceptron and Moving Average Model

In order to improve the performance of high dimensional time series classification, a High Dimensional Time Series Classification Model (HDTSCM) based on multi-layer perceptron and moving average model is proposed. By constructing a multi-layer perceptron neural network model and applying moving average model in the backward propagation of the network model, it realizes the train of the model and classification of high dimensional time series. Experimental results on 8 UCRArchive datasets show that the classification error rates of HDTSCM are significantly lower than the classification methods of Euclidean distance and dynamic time warping, relatively reduced by 49.76% at most.