Towards Efficient and Scalable Ensemble Machine Learning Models for Lung Cancer Diagnosis

Lung cancer is considered one of the major causes of death worldwide. Various external habits and internal human factors contribute to its development. Early prediction of lung cancer can significantly reduce the risk of death, and existing research focuses on predicting it using advanced Artificial Intelligence (AI) techniques, such as Machine Learning (ML). Most studies aim to classify data into either binary categories (cancer or no cancer) or multi-class categories to determine the stage of cancer. Researchers primarily focus on improving performance for specific scenarios. However, it is essential to design a model capable of performing efficiently across various data labels and scales. With this perspective, we have developed an ensemble ML model using a stacking process that integrates Logistic Regression, Decision Tree, EXtreme Gradient Boosting, and Random Forest classifier namely Stacking LDXR, collectively referred to as Stacking LDXR. The proposed method achieves 98% accuracy, precision, recall, f1 score, and sensitivity in both binary and multi-class dataset and 2% False Positive Rate (FPR), 95% MCC and 95% Cohen Kappa in binary classification and 1% FPR, 97% MCC, and 96% Cohen Kappa in multi-class classification, outperforming existing models. The LDXR model addresses the challenges of both binary and multi-class classification by delivering superior performance in both scenarios. This study sheds light on addressing such challenges in this critical domain.