Comparison of Discretization Approaches for Granular Association Rule Mining

Granular association rule mining is a new relational data mining approach to reveal patterns hidden in multiple tables. Recently, granular association rules have been proposed for cold-start recommendation, where a customer or a product has just entered the system. The current research considers only nominal data. In this paper, we study the impact of discretization approaches on mining semantically richer and stronger rules from numerical data. Specifically, the equal width, the equal frequency, and the k-means approaches are adopted and compared. The setting of interval numbers is a key issue in discretization approaches. Therefore, different settings are compared through experiments on a well-known real life data set. Experimental results show that: 1) discretization is an effective preprocessing technique in mining stronger rules; 2) the appropriate settings of interval numbers are critical to obtaining more rules; 3) the equal frequency approach outperforms the equal width and the k-means approaches; and 4) the recommendation accuracy and the number of recommendations are improved significantly through the discretization approaches.