Farman, MuhammadSadri, KhadijehNisar, Kottakkaran SooppyAmilo, DavidIzuchukwu, Roseline ChineloHafez, Mohamed2026-04-212026-04-2120262666-720710.1016/j.rico.2026.1006892-s2.0-105032921841https://hdl.handle.net/123456789/8965https://doi.org/10.1016/j.rico.2026.100689The research introduces a new system that combines machine learning with fractional-order differential equations to forecast cervical cancer risk and track its progression through vital risk elements. The Hospital Universitario de Caracas in Venezuela provided a dataset of 858 records with 36 features, which Random Forest, XGBoost, Support Vector Machines, and Gradient Boosting models used to achieve an XGBoost AUC of 0.992 for binary classification despite class imbalance, which was handled using Synthetic Minority Over-sampling Technique (SMOTE). The Caputo FDE system with fractional-order 0.6085 received the ML-derived probabilities to simulate five normalized risk factors, which include first sexual intercourse and age, intrauterine device years, sexually transmitted diseases, and hormonal contraceptives during normalized pseudo-time. The optimization process using sequential quadratic programming achieved a weighted mean squared error of 0.0042, which demonstrates precise matching between predictions and actual data. The main results show that ML-FDE probability surfaces improve risk prediction by demonstrating higher cancer risks for women who use hormonal contraceptives for extended periods and have sexually transmitted diseases. The method provides accurate results for both high and low probability cases. The system integrated the predictions through its real-time diagnostic system.eninfo:eu-repo/semantics/openAccessFractional CalculusPreventable DeathsMachine LearningCervical CancerPublic HealthArticle