PREDICTIVE RISK ASSESSMENT OF MUNICIPAL ASSETS USING MACHINE LEARNING: MODELING AGE, DEPRECIATION, AND USAGE FOR MAINTENANCE DECISIONS

Abstract

With aging infrastructure and limited budgets, municipalities must make informed decisions on asset maintenance and modernization. This study presents a machine learning-based model for intelligent risk assessment of municipal property. A real dataset of over 620,000 records on government agency assets was used, including acquisition date, cost, classification, and IT affiliation.

Features were engineered to reflect technical and operational condition—such as asset age, depreciation level, and usage intensity. Assets were classified as “high-risk” or “low-risk” based on logical rules aligned with industry standards. Predictive models were developed using Logistic Regression, Random Forest, and XGBoost. The XGBoost model showed the best performance, minimizing false positives and misses. Feature importance analysis confirmed the significance of depreciation, age, and usage intensity. The results can be integrated into digital asset management platforms to support a shift toward proactive maintenance. The study demonstrates the strong potential of machine learning in managing urban infrastructure effectively.