Performance Analysis of Machine Learning-Based Breast Cancer Detection Algorithms

Abstract

Breast Cancer has now been a threat to the lives of countless women. This growth of breast tissue is metastatic, and therefore grows rapidly, infecting other body parts too. The probability of survival is high only if the tumor is detected in an early stage, the higher the stage, lower are the chances of survival of the patient. The presence of a minor tumor could be missed by the human eye, but the machine learning algorithms scan mammograms deeply and are able to detect even the smallest tumor. This work is a performance analysis of three Supervised Machine Learning Algorithms, namely, Convolutional Neural Networks (CNN), Random Forest (RF), and Support Vector Machine (SVM), on two distinct datasets, i.e., Breast Cancer Wisconsin (Diagnostic) dataset and Breast Histopathology Images dataset. Univariate feature selection methods have been applied to select ten features in Breast Cancer Wisconsin (Diagnostic) dataset, and Wrapper Feature Selection methods have been applied to select three instances containing ten features in the Breast Histopathology Images dataset. The results exhibit that RF is the best suited algorithm for the Breast cancer Wisconsin (Diagnostic) dataset with an accuracy of 98.91%, while CNN is suitable for Breast Histopathology Image Dataset with an accuracy of 92.4%. Further, the effectiveness of this machine learning model is tested using the k-fold cross-validation technique.