Mining both associated and correlated patterns

Abstract

Association mining cannot find such type of patterns, "the conditional probability that a customer purchasing A is likely to also purchase B is not only greater than the given threshold, but also much greater than the probability that a customer purchases only B. In other words, the sale of A can increase the likelihood of the sale of B." Such kind of patterns are both associated and correlated. Therefore, in this paper, we combine association with correlation in the mining process to discover both associated and correlated patterns. A new interesting measure corr-confidence is proposed for rationally evaluating the correlation relationships. This measure not only has proper bounds for effectively evaluating the correlation degree of patterns, but also is suitable for mining long patterns. Our experimental results show that the mining combined association with correlation is quite a valid approach to discovering both associated and correlated patterns. © Springer-Verlag Berlin Heidelberg 2006.