Boosting Cervical Cancer Prediction Leveraging a Hybrid FT-Transformer Model

Cervical cancer remains one of the leading causes of cancer-related deaths in women worldwide. Early and accurate prediction of cervical cancer is critical for timely intervention and improved patient outcomes. This study explores the use of advanced machine learning techniques to predict cervical cancer, leveraging a diverse set of features, including demographics, lifestyle factors, and medical history. Several hybrid deep learning models are proposed, including a novel Hybrid FT-Transformer model that combines FT-Transformer, DepthwiseConv1D, and LSTM, integrated with Explainable AI (XAI) methods such as SHAP and LIME for interpretability. Other models explored include CNN + BiLSTM, AutoEncoder + MLP, and TabNet, an attention-based learning model designed for tabular data. The models are evaluated using K-fold cross-validation and a train/validation/test split of 80-10-10. SMOTE is applied to balance the training dataset. The performance of the models is measured using key metrics such as accuracy, precision, recall, F1-score, and AUC. The results highlight the potential of hybrid deep learning models, particularly the FT-Transformer-based approach, in achieving high accuracy and interpretability, thereby offering a robust solution for cervical cancer prediction. This work underscores the significance of AI-driven techniques in medical diagnostics and provides valuable insights for improving cervical cancer detection and patient care.

Keywords:

Cervical Cancer Prediction, Hybrid FT-Transformer Model, SHAP, LIME, CNN + BiLSTM, AutoEncoder + MLP, TabNet, Machine Learning, Deep Learning, Classification, Medical Diagnostics, SMOTE, Accuracy, Precision, Recall, F1-score, AUC, Explainable AI (XAI).