Visualization of Frequent Itemsets with Nested Circular Layout and Bundling Algorithm

Abstract

Frequent itemset mining is one of the major data mining issues. Once generated by algorithms, the itemsets can be automatically processed, for instance to extract association rules. They can also be explored with visual tools, in order to analyze the emerging patterns. Graphical itemsets representation is a convenient way to obtain an overview of the global interaction structure. However, when the complexity of the database increases, the network may become unreadable. In this paper, we propose to display itemsets on concentric circles, each one being organized to lower the intricacy of the graph through an optimization process. Thanks to a graph bundling algorithm, we finally obtain a compact representation of a large set of itemsets that is easier to exploit. Colors accumulation and interaction operators facilitate the exploration of the new bundle graph and to illustrate how much an itemset is supported by the data.