Designing a Revenue Blueprint for JKN: A Hybrid Time-Series and AI Forecasting Approach

257 - 268

Abstract

Indonesia’s National Health Insurance (JKN) system faces significant sustainability challenges driven by regional economic heterogeneity and fluctuating payment compliance, particularly within the informal sector. Traditional linear forecasting models often fail to capture the complex, nonlinear interactions between revenue and broader macroeconomic drivers. This study aims to propose a Hybrid Time Series-AI framework for monthly JKN contribution revenue forecasting to enhance financial resilience. The methodology employs an additive decomposition approach, integrating Autoregressive Integrated Moving Average (ARIMA) to capture linear trends and Long Short-Term Memory (LSTM) networks to model nonlinear residuals and exogenous variables. The research utilizes a spatially enriched dataset (2020–2025) incorporating national macroeconomics and regional indicators such as provincial minimum wages (UMP) and Gross Regional Domestic Product (GRDP). Results reveal that JKN revenue is highly elastic to UMP (r > 0.8) and regional economic stability. The hybrid model, validated through rolling forecasts, yielded a Weighted Absolute Percentage Error (WAPE) of 45.18% during the volatile 2024 period, effectively capturing structural shocks that standard models overlook. These findings are translated into a “Blueprint Revenue JKN Masa Depan,” providing BPJS Kesehatan with a granular tool for adaptive liquidity buffering, region-specific risk management, and long-term fiscal planning.