Adapting k-means for graph clustering

Abstract We propose two new algorithms for clustering graphs and networks. The first, called K?algorithm , is derived directly from the k -means algorithm. It applies similar iterative local optimization but without need to calculate means. inherits properties of in terms both good capability tendency get stuck at a optimum. second algorithm, M-algorithm gradually improves on results K -algorithm find potentially better optima. repeatedly merges splits random clusters tunes with -algorithm. Both are general sense that they can be used different cost functions. consider conductance function also introduce functions, inverse internal weight mean . According our experiments, M outperforms eight other state-of-the-art methods. perform case study by analyzing disease co-occurrence network, which demonstrate usefulness an important real-life application.