Unemployment rate forecasting in Indonesia using macroeconomic indicators with a machine learning approach

Forecasting unemployment is crucial for economic stability and effective policymaking. While traditional econometric models are widely used, they often struggle with the complex, non-linear patterns found in macroeconomic data. This study addresses this gap by developing and evaluating a machine learning framework to forecast Indonesia’s unemployment rate. Using annual time series data from the World Bank spanning 1970 to 2023, this research investigates the predictive power of key macroeconomic indicators: GDP growth, higher education attainment, inflation, and foreign direct investment (FDI). A comparative analysis of eight machine learning algorithms—including linear, ensemble, and non-linear models—was conducted. The Gradient Boosting Regressor emerged as the most accurate model, achieving an R2 of 99% on the training data and the lowest error on hold-out and cross-validation tests. Feature importance analysis revealed that higher education was the most influential predictor of unemployment. For forecasting the 2024–2028 period, a hybrid ARIMA-ML approach was implemented. Future values of the macroeconomic indicators were first projected using ARIMA models, and these projections were then used as inputs for the trained gradient boosting model. The results suggest moderate fluctuations in Indonesia’s unemployment rate over the next five years. This study validates a robust machine learning methodology for economic forecasting in Indonesia, offering policymakers a reliable tool and providing empirical insights into the key drivers of the nation’s labour market.