Main Article Content
Abstract
This research focuses on improving breast cancer classification through a combination of Random Forest and Particle Swarm Optimization (PSO) algorithms. Being the most common cancer among women worldwide, breast cancer requires an effective diagnostic screening method. Traditional methods such as manual examination and X-ray imaging are time-consuming and prone to errors. This research applies machine learning techniques, specifically Random Forest, for image classification based on mammograms. The methodology involves data collection, image preprocessing (including image resize, grayscale, and image segmentation using Sobel Edge Detection and Adaptive Thresholding), feature extraction via Local Binary Pattern (LBP), and classification via Random Forest optimized with PSO. PSO helps to identify the optimal hyperparameters and improves the accuracy of the Random Forest model. Model evaluation is done using confusion matrix which includes accuracy, precision, and recall values. The testing experiment showed that the PSO-optimized Random Forest model achieved an accuracy of 88.37%, outperforming the standard Random Forest model which achieved 86.05%. This shows that PSO significantly improves classification accuracy. This research contributes to the development of an easy-to-use diagnostic tool to assist specialists in accurately identifying breast cancer stages, and suggests future investigations should incorporate additional machine learning algorithms and utilize higher-standard DICOM images to improve training and testing data.