A Propose on a Varied Dual Match Method for Subsequence Queries

摘要

Time-series data has been widely used in many databases applications, such as data mining and data ware-housing. One of the main tasks in handling time-series data is to find subsequences matches similar to a given query sequence. The state-of-the-art methods to find subsequences matches in time-series data leave many remained data sequence not compared for determining candidates and thus, produce many false alarms. In this paper, we propose a varied dual match method for subsequence queries. In our proposed method, we compare the remained data sequence for determining candidates, and calculate the distance between data subsequence and query sequence. Hence, our proposed method reduces the number of candidates and false alarms. Through the experiments with synthetic data and query sets, we show that our proposed method reduces the number of candidates by up to several times comparing to the state-of-the-art methods.