A new maximal-clique-based clustering approach using k-nearest-neighbor

Clustering is an unsupervised classification technique for data analysis that iswidely used in many fields. In clustering techniques, finding the similarity betweendata points or objects’ features plays a central role in community detection. Mostexisting approaches handle this task based solely on the distance between given data.Besides, in hard clustering methods, knowing the number of final clusters before theimplementing algorithm questions the concept of unsupervised in these methods. Toget more coherent clusters whose cores are well-connected sub-graphs, the presentstudy suggests a new maximal-clique-based clustering approach where the similaritynetwork is generated based on an attribute-structural similarity relation between datapoints. The proposed model detects the communities without predefining the numberof clusters which provides a more flexible framework. The experimental results showthe accuracy and capability of the proposed algorithm for data clustering.

Keywords:

Fuzzy similarity , k-Nearest-Neighbor digraph , Clustering , Maximal clique