The article compares the forecasting abilities of traditional econometric methods (Tobit and Heckit) with machine learning techniques (SVR, RF, GBRT) using sample data. The findings show that Tobit performs slightly better in out-of-sample predictions compared to machine learning methods, while RF excels in in-sample predictions. For Heckit, GBRT performs best in RMSE, and Heckit is slightly better in MAE for out-of-sample predictions. Overall, traditional econometric methods focus on accurate parameter estimation, while machine learning techniques prioritize predictive power, but both show similar performance in forecasting. This comparison helps understand the strengths and weaknesses of both approaches, potentially aiding in learning for both methods.