Breast Cancer Detection Using Classifiers

Abstract

Detection of breast cancer is the major phase in Cancer Diagnosis. So, classifiers with higher accuracy are always superior. A classifier already carrying high accuracy and then leading it to higher accuracy offers very less chance to a patient to be wrongly classified. This book involves this kind of classifiers i.e. Naïve Bayes, J48 algorithms along with their performance evaluating criteria. To check up, java based WEKA classification is done with similar dataset & similar feature selection. Modification in typical Naïve Bayes introducing Multivariate Gaussian distribution results in higher accuracy in the thesis work. Fusion in predicted results of the Naïve Bayes & J48 introduces a new algorithm to detect both classifiers’ wrong predictions. So, counting a patient cancerous only in the case of two classifiers saying a patient cancerous leads to poor accuracy overall but more precise prediction. Our thesis work proposed for the last two algorithms while leaving a good overview of breast cancer detection through the Machine Learning Classifiers.