Comparative Study of Different Machine Learning Models for Breast Cancer Diagnosis

Abstract

One of the primary causes of disease among women is breast cancer. According to a survey, one out of every twenty-eight women in India is susceptible to breast cancer. Early detection and avoidance methods can reduce the prospects of death considerably. Three methods are available for breast cancer detection—mammography, FNA, surgical biopsy. The sensitivity of mammography varies from 65 to 79% that of FNA varies from 65 to 85%. Surgical biopsy is close to 100% but is tedious and expensive. Diagnosis of breast cancer is a noteworthy area of data mining study. In our work, a comparison among the various classifiers like logistic regression, KNN classifier, naïve Bayes, state vector machines, decision trees, random forest classifier on Wisconsin diagnostic breast cancer (WDBC) dataset based on accuracy and confusion matrix is offered. In addition, it presents a combination of different classifiers by means of stacked generalization method. Different dimensionality reduction methods like least discriminant analysis and principal component analysis have been used on the dataset. All models are programmed in Python 3 using Anaconda environment.