An Ensemble Strategy- Combining XGBoost and MLP for Cervical Cancer Risk Prediction

This research work aims to develop and evaluate an ensemble model for cervical cancer risk prediction, leveraging the strengths of XGBoost and Multi-layer Perceptron (MLP) to enhance predictive accuracy and identify the most significant features for accurate classification, ultimately providing a robust tool for early cervical cancer detection. This study employs an ensemble learning approach that combines the powerful XGBoost classifier with the MLP model, with feature selection performed using Modified Recursive Feature Elimination (MRFE) to categorize the most relevant predictors. The ensemble model is trained on a dataset comprising essential risk variables for cervical cancer, and its efficacy is assessed utilizing standard classification metrics and ROC analysis. The suggested ensemble model exhibits notable enhancements in predictive performance, attaining an accuracy of 96.97%, a sensitivity of 100%, a specificity of 94.23%, and an F1 score of 0.9703. The ROC curve study demonstrates an area under the curve (AUC) of 0.97, signifying exceptional model performance. In comparison to individual classifiers, the ensemble model offers higher predictive reliability and robustness, confirming its effectiveness in cervical cancer risk classification. The integration of XGBoost and MLP in an ensemble framework, coupled with efficient feature selection via RFE, presents a novel approach for cervical cancer risk prediction. The ensemble model’s superior performance, as evidenced by the high accuracy, sensitivity, specificity, and AUC, makes it a support system for medical decision support in the early detection of cervical cancer.