An improvement of Disk Aware Discord Discovery Algorithm for Discovering Time Series Discord

A time series is a sequence of data points where each point represents a value at a given point in time. The problem of discovering discord in time series has received a lot of attention lately. Time series discord is a subsequence of a long time series which is the most different from all the rest of the time series subsequences. In this work, we propose an improvement of Disk Aware Discord Discovery (DADD) algorithm for time series discord discovery. The improvement is based on symbolic aggregate approximation method associated with a hash bucket structure to speed up the selection for discord candidates. The experimental results showed that our improved algorithm outperforms the original method, Disk Aware Discord Discovery, in terms of runtime while the accuracy is the same.