Median-difference window subseries score for contextual anomaly on time series

Anomaly detection in time series is one of exciting topics in data mining. The aim is to find a data point which is different from the majority, called an anomaly. In this paper, a novel anomaly score called Median-Difference Window subseries Score (MDWS) is proposed with its algorithm and the recommended window size for detecting the contextual anomalies on time series data. It is computed as the subtraction of the middle point with the median of all data points within the current window. The proposed MDWS algorithm is implemented as the median-update of the current window subseries to maintain the linear time complexity. Two anomaly thresholds are set as the mean plus/minus three standard deviation for extracting the anomalies. Furthermore, the suitable window size for detecting anomalies is investigated and suggests that it should be smaller than the seasonal period. The experimental results show that the MDWS has the highest accuracy performance on the benchmark datasets from Yahoo comparing with others existing anomaly detection methods.