Time series forecasting based on similarity-based clustering algorithm

This paper presents a new method to forecast the behavior of time series based on similarity-based clustering algorithm. First, since dynamic time warping (DTW) distance has been proven to be one of the most accurate distance measures for time series clustering, this paper proposes a modified fuzzy c-means based on dynamic time warping (DTW-FCM). Then, DTW-FCM is employed to cluster the time series data. Finally, the phase of forecasting is applied by using the information provided by this clustering. Results from Taiwan Stock Exchange Capitalization Weighted Stock Index (TAIEX) are reported and the performance of the proposed method is compared with that of published techniques showing an improvement in the prediction. © 2012 ICIC International.