An ensemble classification based approach for breast cancer prediction

Abstract

Breast cancer is the second major reason for deaths in women. Early detection of the breast cancer and receiving the appropriate treatment can reduce the death rates as survival becomes hard in the higher stages of the tumor growth. Application of machine learning in healthcare play a key role in aiding the clinical experts to detect the disease at a early stage and perform precise assessment. This paper proposes a pattern recognition methodology that uses breast cancer biomarkers as the attributes and ensemble classification approach for accurately detecting the presence of cancer. The proposed method was evaluated for the samples of the breast cancer Coimbra dataset by fusing the decisions of naive Baye's, radial basis function neural network and linear discriminant analysis classifiers based on majority voting rule. The experimental results demonstrated the enhanced performance of the system with fusion of classifiers as compared to the single classifiers.