Discovery of Multi-frequent Patterns Using Directed Graph

Abstract

In this paper, an algorithm is proposed for mining frequent maximal itemsets. Discovering frequent itemsets is the key process in association rule mining. One of the major drawbacks of traditional algorithms is that they work only for items with single frequency per transaction. Proposed algorithm works with multiple frequency of an item per transaction. Proposed algorithm scans base database only once. The proposed algorithm took lesser time to find candidate itemsets. It uses directed graph for candidate itemsets generation. We conducted experiments on two datasets, Mushroom and Chess. Experimental results showed that proposed algorithm can quickly discover maximal frequent itemsets and effectively mine potential association rules.