Fuzzy sequential pattern mining with sliding window constraint

Abstract

Sequential pattern mining is to discover all subsequences that are frequent. The classical sequential pattern mining algorithms do not allow processing of numerical data and require preprocessing of these data into a binary representation, which necessarily leads to a loss of information. Fuzzy sets are used to overcome this problem. In present fuzzy sequential pattern mining algorithms, there isn't any matter of itemset time and sequences are only found based on sequence of happening. In this paper, a novel fuzzy sequential pattern algorithm is proposed with sliding window constraint which permits elements of a pattern to span a set of transactions within a user-specified window. Therefore, loss of useful sequences is prevented in the search process. The proposed algorithm searches for a goal sequence within the defined fuzzy sliding window and the membership degree of sliding window is returned if the goal sequence is found.