Stacking Ensemble Technique to Predict Cervical Cancer Using Hyperparameter Tuning and Feature Selection

Cervical cancer is the primary cause of premature mortality for women in underdeveloped countries, where it causes approximately 85% of cases. Better outcomes from treatment and preventative measures depend on early risk factor assessment. A thorough prediction model based on preliminary testing data and individual medical records is presented in this study. The risks were analyzed using algorithms such as Random Forest (RF), Decision Tree (DT), and AdaBoost. SelectKBest, XGBoost, and Chi-Square are examples of feature selection approaches that improved feature extraction, improving performance and interpretation. The algorithms for machine learning that were used were LR, LGBM, SVM, KNN, and Naive Bayes. To attain the highest accuracy, grid search and cross-validation were used for hyper parameter optimization. Stacking ensembles combined RF, DT, and AdB, achieving 99.443% accuracy. Ensemble approaches, particularly when resolving decreases in features and disparity in class, helped to make clinical decisions and accurately predicted the risk of cervical cancer.