Abstract:

Considering users’ access order, the recommendation approach based on sequential patterns is becoming one hot topic in the field of recommender system. To improve the level of personalization, we propose a recommender algorithm named personalized recommendation based on weighted sequential patterns(PRWSP). We first present a new weighted sequential pattern model, in which the different importance degrees of the items in different sequences are considered. Furthermore, by approximation, the rationale of antimonotonicity in mining weighted sequential patterns is discussed, thus the searching space is reduced. Finally, the measurement metrics of the matching degree of the sequential patterns are defined. Experimental results show that the PRWSP algorithm has higher mining efficiency and recommendation accuracy.

Key words: data mining;weighted sequential pattern;antimonotonicity;recommender algorithm