Abstract

We introduce a new clustering algorithm which is based on the combination of GA and a new technique called split-and-fusion. GA is used to find the initial cluster while split and fusion refines the cluster by continuously breaking apart and merging patterns existing in the cluster. The whole process is repeated until all patterns have been clustered. The algorithm then merges the smallest-sized cluster with other clusters until termination condition is met. In the last step, a heuristic equation is used to evaluate the termination criteria. Experimental results show that the algorithm is accurate in clustering real-world datasets such as Iris and Wine datasets.