Abnormal pattern detection algorithm for time series based on important edge points
SU Jin-q · Computer Engineering and Science · 2014
Based on analyzing thoughts and concepts of edge operator and advantages and disadvantages of existing pattern representation of time series algorithms,combing with edge point method and important point method,a new algorithm,named pattern representation algorithm based on important edge points of time series,is proposed and analyzed.According to the marginalization of observation point,time series is divided into a plurality of sub-segments by extracting the important edge point.Then,the straight line segment that abnormal values is higher,also called abnormal sequence pattern,is found by the analysis of its similarity.The algorithm is analyzed and verified theoretically and in experiments.It is proved that the new algorithm has low complexity,small pattern representation error and can satisfy the pattern representation requirements of mass time series data.