Enhancing Cervical Cancer Prediction with Machine Learning and Explainable AI

Abstract

Cervical cancer, the fourth most prevalent cancer among women globally, poses significant health challenges, especially in low- and middle-income countries where diagnostic resources are scarce. Early detection is crucial to improving survival rates, yet traditional screening methods such as Pap smears and HPV testing face limitations in cost, accessibility, and accuracy. Advances in machine learning and Explainable AI offer promising solutions to enhance diagnostic accuracy and transparency, addressing these challenges. This study aims to develop a robust cervical cancer prediction system using an ensemble of Decision Tree, Gradient Boosting, and XGBoost models. The research integrates systematic data preprocessing, feature selection, hyperparameter tuning, and data balanc- ing techniques to mitigate dataset limitations and class imbalances. Explainability tools, SHAP and LIME, provided transparency into model decisions, ensuring clinical trust and usability. The proposed system achieved an accuracy of 99.57%, with a precision of 94.8%, a recall of 92.5%, and an F1 score of 93.64%. Compared to existing models, this approach demonstrates superior predictive performance and interpretability. Discussions focus on the model’s potential to reduce mortality rates by enabling early diagnosis, particularly in resourceconstrained settings. Future work includes validating the model on diverse pop- ulations and exploring itsintegration into clinical workflows for real-world applicability.