Automated Data Collection and Predictive Budget Analysis for Government Fleet Scheme Management

Government Fleet Scheme Management in Rwanda has historically relied on manual, paper-based processes, resulting in inefficiencies, lack of transparency, and forecasting inaccuracies. This work focuses on the development and evaluation of a web-based platform GFSMIS (Government Fleet Scheme Management Information System) designed to automate scheme workflows and incorporate predictive analytics using machine learning techniques. A linear regression model was trained on historical scheme data (2021/2022 to 2023/2024) to predict annual budget from for beneficiaries and a random forest model was trained on historical data from 2021/2022 to 2025/2026 to predict annual budget for institutions. GFSMIS integrates user role management, digital signature verification, and real-time data visualization through dashboards. The linear regression predictive model achieved a Mean Absolute Error (MAE) of 183,217,212 FRW and a Relative Error of 1.21% for beneficiary schemes, demonstrating high accuracy and the random forest predictive model achieved a MAE of 428,794,619 FRW with a relative error of 2.13% for institution schemes, demonstrating high accuracy as well. The system supports fiscal planning, transparency, and automation for national-level decision-making.