Sequential Pattern Mining Algorithm Based on Interestingness

Abstract

As an important task of data mining, sequential pattern mining is also an important and challenging task in many practical applications. Traditional sequential pattern mining algorithms usually use frequent patterns as interesting sequence patterns, but frequency is not always a good proxy for interestingness This paper proposes a sequential patterns mining algorithm based on interestingness, using the branch and bound method of OPUS to traversal all possible candidate sequences. Algorithm uses bitmap as data structure to reduce storage space and experimental results on synthetic data with known patterns and experimental results confirms the validity of proposed algorithm.