Machine learning–based evaluation of village fund management: enhancing efficiency and accountability

Abstract

This study aimed to develop and validate a machine learning–based governance analytics framework for evaluating the efficiency and accountability of Village Fund Allocation (ADD) management. A quantitative explanatory design was employed using panel data from 83 villages in Kabupaten Buol during 2022–2024, comprising 249 village-year observations. Financial and governance indicators were integrated into Random Forest regression and classification models using grouped cross-validation. The regression model achieved a Mean Absolute Error of 7.8% and an R² of 0.82, while the classification model attained an Accuracy of 0.87 and an ROC-AUC of 0.91. Budget deviation, reporting timeliness, transparency, and managerial capacity emerged as the principal determinants of efficiency. The findings demonstrate that governance quality plays a more important role than fiscal magnitude in improving village fund efficiency and support the use of machine learning for risk-based fiscal supervision.