Unsupervised Clustering Strategy Based on Label Propagation

This paper propose three novel approaches for clustering, called LPK-means algorithm, LPK-medoids and LPMK-medoids, based on label propagation algorithm. LPK-means algorithm runs like k-means algorithm, meanwhile LPK-medoids algorithm and LPMK-medoids run like k-medoids algorithm. The three proposed algorithms partition clusters by label propagation. To evaluate the proposed algorithms, we use six UCI real datasets and compare with the results of k-means algorithm and k-medoids algorithm in terms of Rand index, precision, recall and F-score. Experimental results show that the three proposed algorithms perform much better than the k-means algorithm and k-medoid algorithm.