Performance of Shrinkage Methods for Forecasting GDP

Abstract

The purpose of the study is to compare performance of three advanced shrinkage methods namely Ridge Regression, LASSO, and Elastic Net Regression methods, so that to come up with the most appropriate model for forecasting GDP in Indonesia and India. As the hybrid method, The Elastic Net Regression method is a convex combination of  Ridge Regression Penalty and  LASSO Penalty.  Under the MSE criterion, Elastic  Net Regression performs better compared to Ridge, and LASSO methods, thus the Elastic Net  Regression is an appropriate method for forecasting GDP in Indonesia and India.  Finally, based on the fitted elastic net regression models, we observed that the Number of  International  Tourist Arrivals  and  International Tourism Receipts have a large impact on GDP In Indonesia compared to India, only International Tourism Expenditures  has a large impact on GDP in India compared to Indonesia.