Detecting Outlier Patterns in Time Series Data
Aiguo Li · Journal of Guangxi Normal University · 2006
A systematic method for identifying outlier patterns in time series data is proposed.The method has 4 steps:1) segmenting time series data into a subsequence set;2) transforming these subsequences into a feature space;3) generating a pattern set from the subsequence set with a cluster algorithm;4) defining the concept of Outlier Support (OS),and calculating OS of each pattern.Patterns with OS1 are outlier patterns.Experimental results demonstrate that the proposed method can identify outlier pattern efficiently.