Longest Common Subsequence Based Algorithm for Measuring Similarity Between Time Series: A New Approach
Rahim Khan, Mushtaq Ahmad, Muhammad Zakarya · 2013
Abstract: In this paper, a robust algorithm is presented for finding similarity between two time series. The longest common subsequence (LCSS) is calculated by avoiding unnecessary comparisons. A time control st parameter is introduced to prevent it from matching the 1 element of a time series with the last element of nd 2 time series. Its time and space complexity is less than that of dynamic programming based algorithms because it stores only those elements of time series that are part of LCSS. A special feature of LCSS called special longest common subsequence (SLCSS) is also presented. Both algorithms were evaluated and checked over real time data sets and show excellent results for shorter as well as longer time series. Key words: Time series LCSS SLCSS DPA