Classification of Benign and Malignant Breast Tumor Using Machine Learning and Feature Selection Algorithms

Abstract

Breast cancer is among the most frequent kinds of cancer that may be detected early and treated with a high probability of complete recovery before the disease's progression. The only way to save and decrease breast cancer mortality is by early detection, identification, and efficient treatment. The proper categorization of breast tumors is critical in the practice of medical diagnosis. This study builds different breast tumor classification models based on machine learning algorithms. Support Vector Machine (SVM), k-Nearest Neighbours (KNN), and Random Forest (RF) classifiers are used to build a set of breast tumor classification models. Each classifier is evaluated individually before and after applying a set of feature selection methods to a public breast tumor dataset. Moreover, a set of hybrid machine learning models are created using stacking approach. Results show that machine learning algorithms with feature selection techniques can be effectively used to build breast tumor classification models. The highest f-measure value is 97.30%, which obtained by combining SVM classifier with CLAE as feature selection.