Mining Both Positive and Negative Impact-Oriented Sequential Rules from Transactional Data

Traditional sequential pattern mining deals with positive correlation between sequential patterns only, without considering negative relationship between them. In this paper, we present a notion of impact-oriented negative sequential rules , in which the left side is a positive sequential pattern or its negation, and the right side is a predefined outcome or its negation. Impact-oriented negative sequential rules are formally defined to show the impact of sequential patterns on the outcome, and an efficient algorithm is designed to discover both positive and negative impact-oriented sequential rules. Experimental results on both synthetic data and real-life data show the efficiency and effectiveness of the proposed technique.