Currently, chemotherapy drugs are the first-line treatment for lung cancer patients, and evaluating their efficacy is of utmost significance. However, assessing the clinical efficacy of chemotherapy drugs remains a challenging task. In recent years, machine learning, especially artificial intelligence (AI), has emerged as a transformative tool in the field of oncology, capable of integrating multiple clinical and protein biomarkers for more accurate predictions. The study collected clinical data and hematological parameters from 2115 lung cancer patients at Hubei Cancer Hospital. Ten typical machine learning models were selected to predict OS and PFS, including Decision Tree, Random Forest, Logistic Regression, k-Nearest Neighbors, AdaBoost, XGBoost, and CatBoost. The study found that the CatBoost model performed best in predicting 3-year overall survival (OS) and progression-free survival (PFS), with AUCs of 0.97(0.95-0.99) or 0.95(0.92-0.98). Additionally, the study further analyzed the performance of different machine learning models in patient mortality risk stratification. The CatBoost model excelled in distinguishing between high-risk and low-risk patients, which demonstrated outstanding performance in survival rate prediction at various time points, particularly in predicting survival rates at 1 year (0.54, 0.68), 3 years (0.05, 0.27), and 5 years (0.01, 0.07). The results showed that these models performed well in distinguishing high-risk from low-risk patients, especially the CatBoost model. Therefore, we suggest that these models, particularly the CatBoost model, could serve as practical clinical prediction tools to assist clinicians in developing better and more reasonable treatment plans.