Mining top-K association rules

Abstract

Mining association rules is a fundamental data mining task. However, depending on the choice of the parameters (the minimum confidence and minimum support), current algorithms can become very slow and generate an extremely large amount of results or generate too few results, omitting valuable information.This is a serious problem because in practice users have limited resources for analyzing the results and thus are often only interested in discovering a certain amount of results, and fine tuning the parameters is time-consuming.To address this problem, we propose an algorithm to mine the top-k association rules, where k is the number of association rules to be found and is set by the user. The algorithm utilizes a new approach for generating association rules named rule expansions and includes several optimizations. Experimental results show that the algorithm has excellent performance and scalability, and that it is an advantageous alternative to classical association rule mining algorithms when the user want to control the number of rules generated. © 2012 Springer-Verlag.