Community Detection Using Genetic Algorithm and Improved K-Nearest Neighbor Clustering

Abstract

Community detection is gathering nodes on graph network’s in separate groups that is called community. Detected subgraphs on the network represent the communities in the social network graph. One of the most popular graph clustering algorithms which has become popular due to easy implementation and performance is iterative search algorithm but has problems such as sensitivity to the initial amount and trapping into the local optima trap. At present, k-nearest neighbor clustering is used for clustering data. k-nearest neighbor clustering has sensitivity to neighborhood size k which it’s performance severely relies on neighborhood size k. so in this paper first we proposed local mean vector method in k-nearest neighbor clustering for improving and reducing the sensitivity then we use genetic algorithm with improved k-nearest neighbor clustering in order to community detection. Experiments performed on real networks and its subsequent results suggest that the method studied greatly improves the accuracy