Abstract

The diagnosis of gynecological conditions remains a complex medical challenge, encompassing issues from routine pregnancy to disorders like PCOS and endometritis. This research proposes a hybrid Artificial Neural Network (ANN) designed to assist gynecologists in predicting pregnancy success rates by analyzing serum hormone ratios. Unlike standard architectures, this model integrates the Genetic Algorithm (GA) and Bat algorithm within the final hidden layer. The GA is utilized to optimize testing costs, while the Bat algorithm refines training costs. This synergistic metaheuristic approach enhances model performance and predictive precision. The methodology was validated using a dataset of 35,207 patient records from the Hospital University of Jordan (HUJ). The results yielded a classification accuracy of 96.5%, demonstrating the system's robustness in handling diverse clinical profiles, including patients with normal hormonal ranges and those with complications like diabetes or infections. This tool offers significant potential for enhancing clinical decision-making in reproductive medicine.