Forecasting Turkish Exports: A Comparative Study of Statistical, Machine Learning, and Deep Learning Methods

Ü. Yilmaz, Fehime Günbegi

Abstract

Accurate export forecasting is a critical input for macroeconomic policy formulation, trade strategy design, and fiscal planning. This study presents a systematic comparison of eight forecasting models, namely SARIMAX, Prophet, Hybrid-S (SARIMAX+CatBoost), Hybrid-P (Prophet+CatBoost), XGBoost, CatBoost, Gated Recurrent Unit (GRU), and Neural Basis Expansion Analysis for Time Series (N-BEATS), applied to monthly Turkish export data spanning January 2002 to December 2025, comprising 288 observations. Exogenous variables include the Brent crude oil price, the real effective exchange rate, and the industrial production index. Walk-forward expanding-window validation is employed to ensure that each one-step-ahead forecast is generated exclusively from historically available information, thereby eliminating look-ahead bias; hyperparameters are optimized via Optuna with 50 trials; and statistical significance is assessed using Diebold-Mariano tests. Results indicate that the SARIMAX+CatBoost hybrid model achieves the highest predictive accuracy (R² = 0.857, MAPE = 5.67%), marginally outperforming standalone SARIMAX (R² = 0.853, MAPE = 5.72%). Deep learning methods (GRU, N-BEATS) perform substantially below all other models, achieving R² values of 0.44 and 0.39 respectively, consistent with the broader literature finding that neural networks underperform on low-frequency macroeconomic series with structural breaks and limited sample sizes. This study provides the first unified eight-model benchmark for Turkish export forecasting under walk-forward validation, introduces two novel hybrid architectures validated across multiple structural regimes including the COVID-19 shock, and empirically confirms the data-regime hypothesis: statistical and hybrid methods dominate in low-frequency, small-sample, structurally unstable macroeconomic environments.

Source: semanticscholar · PDF

Read the AI summary and key takeaways for traders on WOBR Quant Research.