Machine Learning Face-Off: Performance Analysis for Breast Cancer Detection

Breast cancer (BC) is the most common cancer among women, underscoring the need for accurate and reliable early detection systems. Machine learning (ML) algorithms have shown great promise in healthcare by improving diagnostic methods for various diseases. This study evaluates the performance of several ML classifiers for breast cancer detection by analyzing ultrasound images. The classifiers tested include Random Forest (RF), Decision Tree (DT), Support Vector Machine (SVM), K-Nearest Neighbours (KNN), and Neural Networks Multi-Layer Perceptron(NNMLP). The dataset was split into various training and testing ratios. Experimental results showed that the NNMLP classifier consistently achieved the highest accuracy, particularly in the 90/10%, 70/30%, and 60/40% splits. The SVM classifier also performed well, especially in the 70/30% split. The study concluded that the 70/30% split is the most effective, yielding the highest accuracy and balanced performance across multiple classifiers. Future studies should concentrate on optimizing hyperparameters, exploring ensemble methods, and enhancing the dataset with additional data augmentation to further boost model performance and reliability in breast cancer diagnosis.