Monitoring Range Motif on Streaming Time-Series

Recent IoT-based applications generate time-series in a streaming fashion, and they often require techniques that enable environmental monitoring and event detection from generated time-series. Discovering a range motif, which is a subsequence that repetitively appears the most in a time-series, is a promising approach for satisfying such a requirement. This paper tackles the problem of monitoring a range motif of a streaming time-series under a count-based sliding-window setting. Whenever a window slides, a new subsequence is generated and the oldest subsequence is removed. A straightforward solution for monitoring a range motif is to scan all subsequences in the window while computing their occurring counts measured by a similarity function. Because the main bottleneck is similarity computation, this solution is not efficient. We therefore propose an efficient algorithm, namely SRMM. SRMM is simple and its time complexity basically depends only on the occurring counts of the removed and generated subsequences. Our experiments using four real datasets demonstrate that SRMM scales well and shows better performance than a baseline.