K-Means-Based Deterministic Simulation of Regional Budget Redistribution Policy

Abstract

Fiscal disparity remains a critical challenge in Indonesia's decentralized governance, characterized by an extreme gap in financial capacity between economic growth centers and hundreds of other regions. This study aims to map these inequality patterns and design an objective budget redistribution policy simulation model. Using the State Revenue and Expenditure Budget (APBN) data for the Fiscal Year 2024, this research integrates Data Mining and Deterministic Simulation approaches. The K-Means Clustering algorithm, validated using the Elbow Method, was employed to segment districts/cities, revealing a heavily right-skewed distribution where Central Jakarta acts as an extreme outlier. To address the limitation of static mapping, a Python-based simulation model was developed to test multiple cross-subsidy scenarios (2.5%, 5%, and 7.5%). The results demonstrate that a 5% budget redistribution from the High Cluster significantly elevates the fiscal capacity of lagging regions, providing an estimated additional subsidy of IDR 1.1 Trillion per region. Sensitivity analysis confirms that even a conservative 2.5% cut yields substantial impact without destabilizing donor regions. This study concludes that integrating clustering and What-If Analysis serves as an effective Decision Support System (DSS) for formulating equitable fiscal policies, though implementation must consider political constraints and regional absorption capacity.