Application of Machine Learning Approaches in Health Care Sector to the Diagnosis of Breast Cancer

Abstract

Breast cancer (BC) is a kind of malignant disease that represents the primary reason of women's death around the world, cancer cells form tumors which lead to weakening the functioning of the immune system. If the main risk factors are known and detected correctly, the cure rate becomes higher, and the inappropriate treatments which are the main cause of death will be avoided. Today, several avenues for advancing breast cancer classification research are being studied, in particular to strengthen screening and develop an early diagnosis plan. The purpose of this paper is to approach the unfolding of machine learning techniques in the clinical field to categorize and discriminate patients between malignant and benign groups. Modeling of cytological characteristics based on machine learning is proposed to improve predictive performance. In this work, three proposed algorithms of machine learning techniques have been used for the analyze and classification of Wisconsin breast cancer database, k Nearest Neighbors (k-NN), Naive Bayes (NB) and Support Vector Machine (SVM). We will compare learning metrics of both, using train/test split and cross validation. The obtained results shows that KNN offers the best accuracy (97.07%), NB classifier (94.15 %) and SVM classifier (94.73%).