Enhancing SVM Classification of Breast Cancer Using Dual-Stage PSO Optimization
Pirapong Inthapong, Thiradet Singin, Sayan Kaennakham · 2024
This research investigates the application of six strategies to enhance the Support Vector Machine (SVM) classification for breast cancer detection. The study spans three cases, each employing a dual-stage Particle Swarm Optimization (PSO) for feature selection and SVM penalty parameter optimization. Moreover, the distinct benefits and applications of five PSO variants are also investigated. The cases collectively demonstrate that PSO can significantly refine SVM performance. Specifically, PSO with an adaptive weight (PSO-V4) consistently emerges as an efficient strategy, achieving high accuracy, precision, recall, and F1-scores across cases. The findings reveal that PSO strategies, especially PSO-V4, enhance the predictive power of SVM classifiers while reducing model complexity, making them promising for clinical applications in breast cancer diagnosis.