Prediction of furniture exports in Türkiye using machine learning methods

Abstract

In this study, it was aimed to predict T & uuml;rkiye's furniture exports for the period from January 2010 to May 2025 using machine learning methods based on macroeconomic indicators. The dataset consisted of 10 economic variables, including the furniture industry production index, capacity utilization rate, import volume, total exports, real effective exchange rate, producer price index, money supply, and oil prices, with 185 observations per variable and a total of 2035 data points. During the data preprocessing phase, no missing or outlier values were detected, and the data were divided into 70% training and 30% testing subsets. Decision tree, random forest, and deep learning models were developed using the software RapidMiner Studio, and hyperparameter optimization was performed through the grid search method. Model performances were evaluated using root mean square deviation, mean absolute error, mean absolute percentage error, and R2 metrics. The results indicated that all models predicted furniture exports with high accuracy. The best performance was achieved by the random forest model, with R2 = 0.977 and MAPE = 5.28% in the testing phase. Furthermore, the variable importance analysis based on the random forest model revealed that the furniture industry production index and capacity utilization rate were the most significant determinants of exports, highlighting the production-driven nature of the sector. These findings demonstrate that machine learning methods can be effectively used in forecasting economic indicators and that T & uuml;rkiye's furniture exports can be reliably predicted through data-driven approaches.