Classification of Lung Cancer–Related Data Using Decision Trees: A Risk Factor Interpretability Study

In this study, lung cancer—related individual and environmental risk factors were analyzed using a Decision Tree–based machine learning approach. Two open-access datasets (“survey lung cancer.csv” and “lung-cancer.data”) were utilized. Since the survey dataset does not contain a direct clinical diagnosis label, the “SMOKING” variable was selected as the target class, and the study was designed as a risk-based classification task rather than a disease diagnosis model. Data preprocessing included handling missing values, encoding categorical variables, and feature normalization. The Decision Tree model was implemented in the MATLAB environment and evaluated using accuracy, precision, recall, specificity, F1-score, and AUC metrics. The model achieved an accuracy of 87.1% with a high Recall (Sensitivity) value of 0.918 and an AUC of 0.93, indicating effective identification of individuals carrying dominant lung cancer—related risk factors. These findings demonstrate that Decision Trees provide an interpretable and practical framework for risk-oriented screening and decision-support applications in healthcare.