A improved common principal components based dimension reduction method for multivariate time series analysis

Existing traditional dimension reduction methods for multivariate time series have limitations for principal feature preservation, and have impact on the quality of data mining. Therefore, from the perspective of shape features of data, a novel dimension reduction method of multivariate time series based on improved common principal components was proposed. In training datasets, centers for multi-time series of each category were obtained through the improved DTW Barycenter Averaging method. And then the common principal component analysis of the central time series in each category is carried out. In this way, the dimension of multi-time series can be reduced. The comparative experimental results show that the proposed method can reduce dimension effectively and achieve a good classification effect.