VARIABLE-LENGTH CONSENSUS MOTIF DISCOVERY IN MULTIPLE TIME SERIES

Time series motif discovery is an important research task in the field of data mining, aiming to find the similar and meaningful sub-sequence fragments in the sequences. However, for the mining of variable-length consensus motif in multiple time series, the existing methods cannot efficiently solve this problem. To this end, an efficient and extensible algorithm VCMD is proposed. This algorithm used a novel lower bound pruning method and combined with a strategy based on frequent item pruning optimization to accelerate the variable-length consensus motif discovery. Experimental results on multiple real data sets show that VCMD can significantly improve the time efficiency and the lower bound performance compared with existing methods.