A New Algorithm for Mining Weighted Closed Sequential Pattern

Abstract

Most previous sequential mining algorithms have the following two main drawbacks: On one hand, all sequential patterns are treated uniformly while sequential patterns have different importance. On the other hand, most of the sequence mining algorithms still generate an exponentially large number of sequential patterns when a minimum support is lowered. In this paper, a weighted closed sequential pattern mining algorithm called WCloSpan is proposed. WCloSpan generates fewer but important weighted sequential patterns in large databases. Our main approach is to push the weight constraints into the sequential pattern growth approach while maintaining the downward closure property. Furthermore, the problem of closed sequential pattern is transformed into closed itemset. Thus, pruning strategies of closed itemset can also be used to enhance the mining efficiency. Experimental results show that the algorithm is efficient and effective.