Unemployment Rate Prediction System Using Deep Learning and XGBoost

Abstract

An accurate prediction of the unemployment rate is crucial for policymakers, economists, and business practitioners, as it supports informed decision-making related to economic planning and resource allocation. This study presents a comparative analysis of two leading machine learning techniques, Deep Learning and XGBoost, in forecasting the Open Unemployment Rate (OUR) in Indonesia. The paper utilizes data from Indonesia’s Central Bureau of Statistics, which is used to estimate unemployment and develop two predictive models: a Deep Learning-based model and an XGBoost-based model. The performance of these models is evaluated using several metrics, including Mean Square Error (MSE), Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), and Success Percentage. The results indicate that XGBoost outperforms Deep Learning in predicting the Open Unemployment Rate in Indonesia. While the Deep Learning model tends to capture more complex patterns, XGBoost offers better interpretability during both the training and testing processes. This comparison highlights the strengths and limitations of each approach in the context of unemployment forecasting and provides valuable insights for future studies and practical applications in economic prediction systems.