A K-Motifs discovery approach for large time-series data analysis

Abstract

Motif discovery is a method for finding some previously unknown but frequently appearing patterns in time series. However, the high dimensionality and dynamic uncertainty of time series data lead to the main challenge for searching accuracy and effectiveness. In our paper, we propose a novel k-motifs discovery approach based on the Piecewise Linear Representation and the Skyline index, which is superior to traditional R-tree index. As the experimental results suggest, our approach is more accurate and effective than some other traditional methods.