Mining Faster, Not Harder: A New Criterion for Gradual Pattern Mining

Gradual patterns highlight correlations between different attributes through rules of the form "more/less X, more/less Y" with X and Y being two of the attributes. These patterns represent valuable knowledge for experts. In the literature, numerous methods allow their extraction by relying on a binary representation of patterns. Although some of these methods enable parallel processing, they consume significant resources (CPU time and memory). In this article, we present a criterion for pruning the number of candidate patterns, thus reducing the search space. Through experiments conducted on both real and synthetic datasets, we compared the effect of the proposed criterion on the performance of two gradual pattern extraction algorithms from the literature, GRITE and Paraminer. The results show a significant reduction in execution time.