A fast chromatic correlation clustering algorithm
Jaishri Gothania, Bala Buksh · 2016
Emerging sources of information like social networks, bibliographic data, and interaction network of proteins have complex relations among data objects and need to be processed in different manner than traditional data analysis. Correlation clustering is one such new style of viewing data and analyzing it to detect patterns and clusters. Being new, it is an open field of research with much scope. This paper discusses a heuristic method to solve the problem of Chromatic Correlation Clustering where data objects as nodes of a graph are connected through color-labeled edges representing relations among objects. Particularly, the low probability of the Chromatic Balls (CB) algorithm to output good quality results in all situations is addressed. The proposed heuristic has higher probability of producing better solutions while retaining the speed related advantages of original CB algorithm. Experimental results prove the efficacy of the proposal.