An Adaptive AI Approach for Cervical Cancer Prediction with Explainability

Cervical Cancer remains one of the leading causes of cancer-related deaths worldwide. Cervical cancer cells develop in the cervix of a woman. HPV infection and other factors, including multiple sexual partners, weakened immune system, early sexual activity, birth control, age, family history, and more, are responsible for cervical cancer. Early diagnosis could be an effective solution to prevent this cancer. So, in this study, we have proposed a model to predict cervical cancer combined with ML algorithms. We used EDA techniques, Median Imputation, and removing zero variance and highly correlated features for data preparation. Our model consists of oversampling techniques such as ADASYN for class imbalances in the dataset. The RFE feature selection technique with RF has been used to select the important features that affect the outcome. ML classifiers that have been used in our model are LightGBM, XGBoost, Random Forest, and Logistic Regression. In our study, we have used an XAI technique SHAP (Shapley Additive Explanations), for interpreting machine learning models. Finally, the ML ensemble classifier (Stacking, Voting) has been used to handle classification problems and achieved 98.03% accuracy through Hyperparameter Tuning and the 5-fold Cross-Validation technique. This study is important because it improves early detection of cervical cancer, helping medical professionals better diagnose and make decisions, which can ultimately reduce mortality rates.