An Improved Framework for Outlier Periodic Pattern Detection in Time Series

Periodic pattern detection in time-series is an important data mining task. Detecting the periodicity of outlier patterns might be more important in many sequences than the periodicity of regular, more frequent patterns. Patterns which repeat over a period of time are known as periodic patterns. Outlier Pattern are those which occur unusually or surprisingly. In this paper, I present the development of an enhanced suffix tree-based algorithm capable of detecting the periodicity of outlier patterns in a time series using MAD (Median Absolute Deviation) is presented. An existing algorithm makes use of mean values, which is inefficient. Use of MAD increases the output of these algorithms and gives more accurate information.