An efficient approach for time series join on subsequence correlation using suffix tree

The JOCOR algorithm, which was first introduced by Mueen et at., is the currently exact method for joining two time series to find the most correlated subsequence. Despite the fact that time complexity for JOCOR belongs to O(n
²
lgn), where n is the length of the time series, it becomes inefficieit and unfeasible even for medium-size time series. This is because the JOCOR method is based on a brute-force approach. In this work, we propose an approximation approach with high rate of accuracy, called suffixTree-JOCOR that runs much faster than JOCOR. Our method consists of three main steps. Firstly, we applied SAX to converse two input time series into two strings of characters. Secondly, the longest common substring (LCS) of the two strings was efficiently found by using suffix tree. Finally, we mapped the resulting LCS into the original time series and applied the exact JOCOR algorithm to find the most correlated subsequence. The experiments show that our proposed method is faster than JOCOR while high accuracy is guaranteed.