Forecasting the Open Unemployment Rate in Banten Province Using the FB Prophet Method in Python Programming Language

Abstract

The Open Unemployment Rate is a key indicator in measuring labor market imbalances, reflecting the economic dynamics of a region. Banten Province has consistently ranked among the top three provinces with the highest Open Unemployment Rate in Indonesia over the past decade, indicating structural challenges in employment. To address this issue, a forecasting model is needed to provide accurate predictions that support more effective labor policy planning. This study uses the Prophet method, an additive regression approach developed by Facebook, to forecast the Open Unemployment Rate in Banten Province over the next 10 semesters (February 2025-August 2029). The data used is sourced from the Statistics Indonesia (BPS) for the period 2005-2024, collected every semester (February and August). The model's performance is evaluated using the Mean Absolute Percentage Error (MAPE) as the primary evaluation metric. The results show that the Prophet model effectively captures trend and seasonal patterns. With a MAPE value of 5.3910%, the model demonstrates very good accuracy (MAPE < 10%), making it suitable for medium-term forecasting. The predictions indicate a downward trend in the Open Unemployment Rate in Banten over the next five years. The conclusion of this study suggests that the Prophet model can be a reliable tool for projecting the Open Unemployment Rate and supporting labor policy planning in Banten. Future research is expected to incorporate external factors or use hybrid modeling approaches to improve prediction accuracy.