Machine Learning-Driven Export Forecasting: A Comparative Analysis for MSME Growth

Abstract

Micro, small, and medium enterprises play a fundamental role in economic devel-opment by fostering employment, innovation, and international trade. However, these enterprises face substantial challenges in volatile global trade conditions, necessitating accurate forecasting methodologies for effective strategic planning. This study aims to evaluate and compare traditional time series models and advanced machine learning techniques in predicting export trends. The research employs Double Exponential Smoothing and Autoregressive Integrated Moving Average alongside Support Vector Regression, Random Forest, and Extreme Gradient Boosting to assess forecasting accuracy. Performance metrics including Root Mean Square Error, Mean Square Error, Mean Absolute Error, Mean Absolute Percentage Error, and R-Square are utilized for model evaluation. Results indicate that while traditional time series models provide foundational forecasting insights, they are outperformed by machine learning techniques. Among these, Random Forest demonstrates the highest predictive accuracy and reliability. However, Extreme Gradient Boosting exhibits near-perfect met-rics, necessitating further scrutiny to address potential overfitting. The study empha-sises the necessity of integrating traditional and machine learning methodologies to enhance forecasting precision. These insights are valuable for policymakers, re-searchers, and industry practitioners seeking to strengthen export strategies and sus-tain economic growth.