Hybrid BATGSA: a metaheuristic model for classification of breast cancer data

Abstract

Nature-inspired algorithms have a vast range of applications. One such application is in the field of medical data mining where, major focus is on building models for the classification and prediction of various diseases. Breast cancer has grabbed the interest of numerous researchers because, it is the major killer disease, killing millions of women across the globe. In this paper, we propose a hybrid diagnostic model which is a fusion of Bat algorithm (Bat), Gravitational Search Algorithm (GSA), and Feed-Forward Neural Network (FNN). Here, the potential of the FNN and the advantages of nature inspired algorithms have been exploited to build a hybrid model used for classification of breast cancer data. The proposed model consists of two modules. First, is the training module where the data is properly trained using a feed-forward neural network and the second, is an error minimising module, which is built using Bat and GSA metaheuristic algorithm. The hybrid model minimises the error thus, producing better classification results. The accuracy obtained for Wisconsin Diagnostic Breast Cancer Diagnostic (WBCD) data set is found to be 94.28% and 92.10% for training and testing respectively.