Development of adaptive neuro-fuzzy inference systems and a user interface for bovine abortive diseases prediction.

Abortive bovine diseases engender significant economic and sanitary challenges to dairy livestock. This study addresses the critical need for improved diagnostic tools by developing new predictive models for four abortive pathogens: (i) Bovine Viral Diarrhea virus (BVDV); (ii) Coxiella burnetii; (iii) Chlamydia abortus, and (iv) Toxoplasma gondii. An advanced hybrid metaheuristic approach, combining Particle Swarm Optimization (PSO) with an Adaptive Neuro-Fuzzy Inference System (ANFIS), was employed in order to create robust and accurate predictive models using a comprehensive dataset of 375 bovine’s samples, collected from six districts in Algeria. ELISA test was used to detect antibodies anti four abortive pathogens. Ten relevant risk factors, including geographic location, age, breed, and husbandry practices, were considered. The PSO algorithm optimized the ANFIS parameters to enhance prediction accuracy for each disease. Model performance was evaluated using training–validation splitting and K-fold cross-validation. The developed models demonstrated high sensitivity (generally > 89%) and good overall accuracy (mostly 85–90%) in both training and validation datasets. Notably, the study translated these sophisticated models into a user-friendly graphical interface, “Bovine Abortive Diseases Diag 2025”, developed in MATLAB. This interface allows veterinarians and researchers to input risk factors and receive immediate predictions on the infection likelihood for each of the four diseases, bridging the gap between advanced modeling and practical field applications.