<![CDATA[This study aims to evaluate the performance of the Bootstrap Aggregating (Bagging) method applied to Multivariate Adaptive Regression Splines (MARS) in improving the predictive ability of biresponse models, compared to biresponse MARS models without bagging, using a case study of inflation and the rupiah exchange rate in Indonesia. Inflation and exchange rates are important macroeconomic indicators that are interrelated and play a crucial role in maintaining economic stability; therefore, a prediction model capable of accurately capturing simultaneous relationships and nonlinear patterns is required. The contribution of this study lies in the application of a biresponse nonparametric regression framework based on bagging to simultaneously model biresponse variables, which has rarely been explored in previous research that generally focuses on a single-response approach. The biresponse approach is used to accommodate the interrelationship between response variables, while the bagging procedure is implemented through a bootstrap technique with several replication scenarios to reduce prediction variance and improve model stability. The final prediction is obtained by averaging the results from all bootstrap models formed. The results of this study indicate that the application of Bagging MARS with 100 replications can significantly improve model performance, as shown by a decrease in the RMSE value from 132.40 to 92.08 and MAE from 70.71 to 52.12, as well as an increase in the R² value from 0.9997096 to 0.9998597. These findings indicate that the integration of the bootstrap technique in the Bagging MARS approach is effective in reducing model variability and producing more stable predictions. Practically, the Bagging MARS method has the potential to be used as an alternative in modeling interrelated macroeconomic indicators with nonlinear characteristics.]]>
