Community Detection Using Revised Medoid-Shift Based on KNN

Community detection becomes an important problem with the booming of social networks. The Medoid-Shift algorithm preserves benefits Mean-Shift and can be applied to problems based on distance matrix, such as community detection. One drawback is that there may no data points within neighborhood region defined by a parameter. To deal problem, new called Revised (RMS) proposed. During process finding next medoid, RMS KNN, while original Since KNN more stable than one parameter in terms number neighborhood, converge smoothly. tested two kinds datasets including known ground truth partition without respectively. experiment results show proposed generally produces better some state-of-the-art together most classic algorithms different datasets.