Abstract
Breast cancer remains a leading cause of mortality among women globally. Early diagnosis and precise classification are essential for treatment and improving survival rates. This research introduces a novel algorithm: the Parallel Enzyme Action Optimizer (PEAO). PEAO introduces three key innovations: a parallel strategy to enhance global exploration, a multi-strategy communication mechanism (optimal replacement, optimal mean replacement, and circular optimal replacement) to improve information exchange, and the Lévy flight strategy to avoid local optima. The proposed algorithm was applied to optimize the hyperparameters of a BiLSTM network for breast cancer diagnosis. Experimental results on the Breast Cancer Wisconsin Diagnostic (BCWD) and Wisconsin Breast Cancer (WBC) datasets demonstrate that the PEAO-BiLSTM model significantly improves classification accuracy, precision, recall, and F1-score compared with recent meta-heuristic algorithms and other hyperparameter optimization methods. These results confirm that the proposed approach provides a robust and reliable diagnostic framework, contributing to the development of intelligent, data-driven systems for clinical decision support in breast cancer diagnosis.