Closeness Preference – A new interestingness measure for sequential rules mining

Abstract

The time-interval between the antecedent and the consequent of a sequential rule can be considered as an important aspect in sequential rules interest. For example, in web logs analysis, the end-user can be interested in predicting the next page that will be visited by an internet surfer based on a history of visited pages. A Closeness Preference measure is proposed to favour the sequential rules with close itemsets based on user time-preference in a post-processing step. We illustrate the interest of the Closeness Preference measure with two real datasets (web logs data and activities of daily living data) for first, a predictive task and second, a descriptive one. Both of them show that Closeness Preference measure is helpful to find small and efficient sets of simple sequential rules.