Metaheuristic Feature Selection Optimization and Deep Learning for Breast Cancer Diagnosis

Breast cancer remains a leading cause of global mortality, emphasizing the need for accurate early diagnosis. This study proposes a hybrid diagnostic approach that combines metaheuristic feature selection with deep learning to improve classification performance. Using the Breast Cancer Wisconsin dataset, two metaheuristics techniques, Ant Colony Optimization (ACO) and Particle Swarm Optimization (PSO) were evaluated for selecting informative features. PSO achieved superior results, selecting 11 out of 30 features and attaining 100% accuracy, outperforming ACO's 98.2%. Feature subsets were evaluated using a Random Forest classifier to guide selection. These optimized features were then used to train deep learning classification models, including Convolutional Neural Network (CNN), Gated Recurrent Unit (GRU), and hybrid CNN-GRU. The CNN-GRU achieved the best performance with 99.1% accuracy and near-perfect precision, recall, and F1 score. This integration of metaheuristics and deep learning demonstrates a promising framework for robust, high-accuracy breast cancer diagnosis. It offers a scalable, intelligent approach for managing high-dimensional biomedical data, contributing to early disease detection and addressing critical public health challenges.