INTEGRATION OF DATA SCIENCE AND AI ENGINEERING FOR CLUSTERING AND FORECASTING OF WEST JAVA REGIONAL BUDGET 2015–2024

Abstract

The management of regional revenue and expenditure budgets (APBD) plays a critical role in supporting local economic development. However, conventional evaluations often fail to capture complex patterns and long-term fiscal dynamics. This study integrates Data Science and AI Engineering approaches to analyze APBD data of municipalities and regencies in West Java Province for the 2015–2024 period. The workflow begins with data preparation, including cleaning, normalization, and the construction of derived variables such as year-on-year growth, the ratio of personnel to capital expenditure, and surplus/deficit status. Exploratory Data Analysis (EDA) was conducted through trend visualization and distribution analysis across regions. Clustering was carried out using the K-Means algorithm and compared with alternative methods such as DBSCAN and hierarchical clustering, evaluated by Silhouette Score and Davies-Bouldin Index. For forecasting, the study employed time-series models ranging from ARIMA and Prophet to advanced LSTM, with accuracy measured by RMSE and MAPE. The findings reveal substantial disparities across regions, such as clusters of high-growth yet recurrent deficit areas, as well as fiscally stable but small-capacity regions. Forecasts for 2025–2026 provide projections of revenue and expenditure, which serve as evidence-based guidance for regional fiscal planning. The main contribution of this research lies in offering a data-driven framework that not only explains historical fiscal performance but also delivers actionable policy recommendations for local governments in West Java.