On a visual frequent itemset mining

Given a large, dense transaction database, generating interesting frequent patterns in a user friendly manner remains as an important issue in data mining. It is because the minimum support, the most popular statistical significance measurement, is not capable of reflecting the domain user's interest. This paper presents visual frequent itemset mining (VFIM) as an alternative to the traditional apriori-like frequent itemset mining. VFIM pushes the domain user's cognitive power into the data mining process. To this end, a formal visual data mining model is proposed and a prototype of the model is created. The effectiveness of the proposed model is demonstrated by showing that VFIM generates frequent patterns, by means of user interaction, that are compatible with those generated by traditional apriori-like algorithms without executing them.