Mining Frequent Coherent Patterns from Weighted Graphs

Abstract

Current research on network analysis; such as community detection, pattern mining and many other graph mining application mostly focus on large social or biological networks. Such experiments may find interesting patterns that helps us to understand the unknown relationship within the network. Sometimes the size of the input networks are so big that it needs an efficient algorithm to overcome the time and space complexities. In this paper we modify an existing algorithm that finds the maximal patterns from a set of input networks. Maximal patterns are those patterns that are not part of any frequent patterns. We introduce a new relational attributes to our algorithm from the input networks, we call them the edge attributes. We have tested our algorithm on a co-author relationship database; and after analyzing we have found some interesting characteristics of the input dataset.