Stacked Multi-Modal Ensemble Classification Approach Using Deep Learning for Lung Cancer Prognosis

Lung cancer is one of the deadliest forms of cancer, with prognosis being life threatening for guiding treatment and refining patient outcomes. Traditional single-modal approaches often lack the comprehensive perspective needed to predict the complex progression of lung cancer accurately. This study presents a novel stacked multi-modal ensemble classification approach using a voting classifier to enhance lung cancer prognosis. The process begins with comprehensive data preprocessing, involving normalization and feature extraction. Various base models, such as Decision Trees (DT), Random Forest (RF), Support Vector Machines (SVM), Convolutional Neural Networks (CNN), Gradient Boosting Machines (GBM), and Logistic Regression (LR), are trained on the processed data. Predictions from these base models serve as input features for a meta-learner in a stacking ensemble framework. Both hard voting (majority voting) and soft voting (probability averaging) strategies are employed to combine the predictions from the base models. The performance of the ensemble model is assessed using metrics such as accuracy, precision, recall, F1-score, ROC-AUC, and confusion matrices.