Bio-inspired computational approaches for breast cancer cluster analysis

Abstract

Breast cancer is the second most widespread disease throughout the world and the maximum incessant cause of death. The possibilities of survival are higher, if it is diagnosed in early phases. Moreover, the lack of awareness, treatment facilities and proactive measures expand the risks of survival. Cluster analysis is a statistical practice that categorizes observations into like sets or groups. The usage of cluster exploration offers a complex challenge as it entails numerous practical selections that define the superiority of a cluster solution. This paper highlights the application of cluster analysis inBreast Cancer dataset with the help of evolutionary approaches. Various evolutionary algorithms like genetic, differential evolution and particle swarm optimization are considered to overcome the problem of local maxima. The work proposes three evolutionary algorithm based techniques named WCGA,WCPSO and WCDE to perform clustering of breast cancer data and evaluate their effectiveness based on clustering validity measure (DBIndex), computation time and in terms of classification parameters. The results show that WCDE outperformed WCGA and WCPSO in terms of DB Index.