Boosting-genetic clustering algorithm

K-means is one of the most popular techniques for clustering problem. However, the quality of resulting clusters heavily depends on the selection of initial centroids and may converge to a local optimum rather than global optimum. Genetic algorithm (GA) has been proposed by many researchers to solve a global solution for clustering problem. Even though GA yields higher accuracy result, it is only practical for small datasets. Clustering large datasets with GA is extremely slow or even impossible. In this paper, we proposed a new clustering algorithm, called Boosting-Genetic Clustering Algorithm (BGCA). Inspired by boosting algorithms, the BGCA algorithm combines multiple clustering results on a small number of specially selected samples and iteratively improves the accuracy of the inconsistent regions of data points. Experimental evaluation shows that BGCA yields a higher accuracy than traditional k-means and is very efficient for clustering large datasets.