Stacked Ensemble Learning Framework for Lung Disease Classification

Lung diseases are increasingly prevalent in developing countries, primarily due to smoking habits, rising pollution levels and chronic inflammation. Smoke chemicals like carcinogens, airborne pollutants including particulate matter and toxic gases, contribute to conditions like asthma, chronic obstructive pulmonary disease, and lung cancer. Artificial intelligence and machine learning techniques are now vital in healthcare, particularly for predicting health risks. In this work, we have proposed stacked ensemble learning classifier to detect the lung disease. Stacked ensemble learning technique utilizes more than one machine learning base-models and thus improves predictive model performance. We have applied feature engineering and adaptive synthetic sampling on publicly available dataset called ‘lung cancer survey’ dataset. The performance of our stacked classifier is compared with other models like random forest, support vector classifier, logistic regression, multi-layer perceptron and gradient boosting. It has been demonstrated that our model obtains the greatest test accuracy of 97.2% along with the highest performance metrics value including precision, recall, and F1-score. For the comparison purpose, confusion matrices of various ML models have been presented. A graphical comparison of performance matrices of different models is shown. The stacked ensemble classifier achieves 99% ROC-AUC score that highlights its robustness and reliability