Time Series Analysis of Libyan Imports and the Impact of Macroeconomic Variables: An Integration of Econometrics and Machine Learning (1998–2022)

Abstract

This study aims to analyse the temporal behaviour of Libyan imports during (1998–2022) and to estimate the impact of key macroeconomic variables (GDP growth, oil production, and inflation) on them, while comparing econometric models with machine learning techniques. The study adopts an integrative approach that combines OLS, ARDL, ECM, and Granger causality with Random Forest, ANN, and SVR, using annual time series data (25 observations).

The results reveal a long-run equilibrium relationship (F= 6.824), with oil production being the strongest determinant (; importance = (52.3%) and a high adjustment speed (ECT=0.682-). Machine learning models outperformed econometric models in forecasting, with the ANN achieving the highest accuracy (R² =0.914, MAPE =15.2%). This study is among the first to integrate econometrics with machine learning in the analysis of Libyan imports.