Adaptive optimized clustering for Veterans' Administration Lung Cancer

Lung cancer is considered as the leading reason of death for men and women. The clustering establishes the clusters from dataset of veterans' lung cancer based on a similarity measure to discover a new set of categories. K-Means, which is used in clustering, owns certain drawbacks such as determining number of clusters being used and determining a well-defined centroid. In this paper, a proposed algorithm called PSO-AKMeans, which is used in Veterans' Administration Lung Cancer (VALC) domain, is presented and evaluated. PSO-AKMeans uses the adaptation concept in AK-Means algorithm and particle swarm optimization to optimize total number of clusters and clusters' centroid which are used in clustering process. The experiment result proves that PSO-AKMeans provides better performance, higher accuracy and lower execution time than other traditional clustering algorithms in VALC domain.