A New Algorithm for Time Series Data Mining by Using Rough Set

Abstract

This paper is to apply Rough Set to data mining of time series. Firstly, we process the time series data by attribute selection and similarity sequence search. Secondly, the time series is partitioned into some sets of pattern by Mobile Window Method (MWM) and each pattern is a trend of time series. Thirdly, an information table is made by predicting attributes and targeting attribute in trending variation ratio structure sequence (TVRSS). Then, the original information table is made suitably for rough set to discover knowledge. Finally, the extracting rules can predict the time series behavior in the future. The total process is four steps. In the end, we show some examples to demonstrate our method on the time series data of stock market.