Anomaly detection in time series data using a fuzzy c-means clustering

Detecting incident anomalies within temporal data - time series becomes useful in a variety of applications. In this paper, anomalies in time series are divided into two categories, namely amplitude anomalies and shape anomalies. A unified framework supporting the detection of both types of anomalies is introduced. A fuzzy clustering is employed to reveal the available structure within time series and a reconstruction criterion is used to assign an anomaly score to each subsequence. In the case of detecting anomalies in amplitude, the original representation of time series is used, while for detecting anomalies in shape an autocorrelation representation of time series to capture shape information is employed. Experimental studies concerning two real-world data sets are reported.