Optimal clustering method based on genetic algorithm

Clustering methods divide the dataset into groups called clusters such that the objects in the same cluster are more similar and objects in the different clusters are dissimilar. Clustering algorithms can be hierarchical or partitional. Partitional clustering methods decompose the dataset into set of disjoint clusters. Most partitional approaches assume that the number of clusters are known a priori. Moreover, they are sensitive to initialization. Hierarchical clustering methods produce a complete sequence of clustering solutions, either from singleton clusters to a cluster including all individuals or vice versa. Hierarchical clustering can be represented by help of a dendrogram that can be cut at different levels to obtain different number of clusters of corresponding granularities. If dataset has large multilevel hierarchies then it becomes difficult to determine optimal clustering by cutting the dendrogram at every level and validating clusters obtained for each level. Genetic Algorithms (GAs) have proven to be a promising technique for solving complex optimization problems. In this paper, we propose an Optimal Clustering Genetic Algorithm (OCGA) to find optimal number of clusters. The proposed method has been applied on some artificially generated datasets. It has been observed that it took less number of iterations of cluster validation to arrive at optimal number of clusters.