A Comparative Assessment of Machine Learning Algorithms for Detecting and Diagnosing Breast Cancer

Abstract

The principal goal of this study was to explore machine-learning techniques deployed for the early detection of breast cancer in the United States. Specifically, three algorithms were trained on a breast cancer dataset: Decision Tree, Random Forest, and Linear Regression. Each model was further evaluated for its performance, to ascertain the best model. Upon review, the Random Forest provided higher classification performance. It was postulated that the Random Forest offered higher accuracy models on the test data because Decision Trees and Linear Regression require more extensive data for them to be more precise in making high-precision predictions. Out of all the models, the Random Forest provided suitable accuracy on test data. Therefore, in this research scope, Random Forest was the most successful and proved effective in accurately identifying breast cancer malignancies. In that light, the proposed random forest can benefit healthcare organizations by facilitating in detection of breast cancer disease by identifying patients in high-risk groups at an early and more treatable stage of disease for improved outcomes and lower healthcare costs. Besides, Random Forest models can assist in identifying high-risk patients in advance for prompt treatment. In that regard, such detection saves lives and decreases long-term healthcare costs for the US government.