Use of Machine Learning Models to Predict Breast Cancer
REU Scholar: Edem Ammamoo (Alcorn State University)
Advisor: Dr. Joshua Young | Mentors: Daniel Mottern and Mo Li
We developed machine learning models to diagnose breast cancer using two datasets: numerical Fine Needle Aspirate data and ultrasound images. For numerical data, we used Logistic Regression, Random Forest, and XGBoost Classifier, achieving 98% accuracy with Logistic Regression. Recursive Feature Elimination identified worst radius, perimeter, and concave points as key features. For image data, we built convolutional neural networks targeting 80% accuracy. Different balancing techniques were tested, with Borderline SMOTE 1 optimal for Random Forest and SMOTE ENN for XGBoost.