A Research on Dimension Reduction Method of Time Series Based on Trend Division

Abstract

The characteristics of high dimension, complexity and multi granularity of financial time series make it difficult to deal with effectively. In order to solve the problem that the commonly used dimensionality reduction methods cannot reduce the dimensionality of time series with different granularity at the same time, in this paper, a method for dimensionality reduction of time series based on trend division is proposed. This method extracts the extreme value points of time series, identifies the important points in time series quickly and accurately, and compresses them. Experimental results show that, compared with the discrete Fourier transform and wavelet transform, the proposed method can effectively process data of different granularity and different trends on the basis of fully preserving the original information of time series. Moreover, the time complexity is low, the operation is easy, and the proposed method can provide decision support for high-frequency stock trading at the actual level.

Keywords

• data dimensionality reduction
• Fourier transform
• time series
• trend division
• wavelet transform