A fuzzy Markovian model for graph-based clustering

Abstract

Graph works well as a random walk model which is a key concept in clustering. The Markov chain model mainly simulates flow in a graph. This paper proposes a random walk-based Markov chain clustering model on bipartite graph. A randomly generated point-based bipartite graph from the spatial data points has been created for this purpose. Proximity among data points has been computed based on the novel fuzzy correlation method and dense subgraphs are formed for random walk to work. Standard Markov flow-based clustering has been applied on the final similarity matrix. The model has been tested with standard datasets and the result has been compared with that of other benchmark models.