Finding time series discord based on bit representation clustering
Knowledge-Based Systems
Hi-index | 0.00 |
In many domains, such as meteorology, space physics, geography, multimedia, and economics, there are a lot of time series data. People often use similarity matching to deal with time series data in practical application. DTW (Dynamic Time Warping) can provide the stretch ability along the time axis. Therefore, it becomes one of several popular similarity matching algorithms. But DTW incurs a heavy computation cost. Although several improved methods of DTW have been proposed in some literatures, it still can not meet the actual needs. This paper proposes FSMBO (Fast time series Similarity Matching based on Bit Operation). FSMBO algorithm efficiently prunes a significant number of the matching candidates, which leads to a direct reduction in the matching cost. Experiments reveal that FSMBO is significantly faster than the popular methods.