Machine Learning Methods for Breast Cancer Diagnosis

Abstract

As many modern diseases begin to surface especially as of late, such as the Ebola and COVID-19 epidemics, scientists have begun developing new and innovative tactics to combat them. While new medicine and vaccines may be developed, one area that needs special attention is the diagnosis of diseases – this is because without a proper and speedy diagnosis, scientists wouldn’t be able to detect diseases, rendering treatment ineffective. Scientists have begun using machine learning algorithms to help ensure an accurate and speedy diagnosis. One specific disease that has seen frequent testing around machine learning diagnosis is breast cancer. Breast cancer is one of the deadliest and common cancers around the world for women, and due to its effects, the doctrine of speed in diagnosis is essential. This study will attempt to find out, out of three machine learning algorithms (neural networks, logistic regression and K-nearest neighbours), which one is the most effective at diagnosing breast cancer using the Wisconsin Breast Cancer Dataset. Results suggest that neural networks perform the best in diagnosing breast cancer, however only by a small margin compared to other results.