Forecasting Youth Unemployment in Malaysia for Ages 15 to 30 Using ARIMA and Prophet Models

Abstract

Accurate forecasting of youth unemployment is essential for effective economic planning and policymaking in Malaysia. However, selecting an appropriate forecasting model remains a challenge, particularly when youth unemployment data exhibit relatively stable trends. This study aims to evaluate and compare the forecasting performance of the Autoregressive Integrated Moving Average (ARIMA) model and the Prophet model in predicting youth unemployment in Malaysia. Monthly historical data were analysed using both models, and their performance was assessed on training and testing datasets using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE). The findings from the testing dataset indicate that the ARIMA (1,1,1) model outperformed the Prophet model, achieving a lower forecasting error with a MAPE of 1.43% compared to 8.92% for Prophet. This suggests that ARIMA provides more accurate forecasts for youth unemployment data characterised by relatively stable trends, while Prophet is more sensitive to complex trend structures that are not prominent in the dataset. In conclusion, the study demonstrates that conventional statistical models such as ARIMA can outperform more recent machine-learning-based approaches when the underlying trends are steady and predictable. The findings provide useful insights for policymakers and government agencies in designing employment programmes and interventions to reduce youth unemployment. Future research may improve forecast accuracy by using longer and higher-frequency datasets, hybrid models, or additional explanatory variables.