A novel enhanced hybrid clinical decision support system for accurate breast cancer prediction

Abstract

Feature selection is one of the crucial data preprocessing techniques for improving the performance of machine learning (ML) models. Recently, metaheuristic feature selection algorithms have become popular because they select optimal features for ML problems. This paper presents three feature selection strategies based on metaheuristic algorithms: Bacterial Foraging (BFOA), Emperor Penguin (EPO), and a hybrid (hBFEPO) combining BFOA and EPO. The baseline algorithms have been investigated for feature selection in other ML tasks, but not for breast cancer classification. A hybrid of these two has been used for the first time. These strategies were initially tested on the COVID-19 dataset. After achieving satisfactory results, these strategies are evaluated on the WDBC Breast Cancer dataset. The performance of our models on WDBC is compared with recent eighteen state-of-the-art studies. The results indicate that the hBFEPO model outperforms other models, achieving 100% precision and specificity, 98.49% accuracy, 95.43% sensitivity, a 95.99% F1-score, and a 99.60% AUC.