Scholarlink

<p>Urban traffic congestion is a persistent challenge in rapidly growing sub-Saharan African cities, where limited sensing infrastructure constrains predictive traffic management. This study presents a data-driven intelligent traffic routing framework for real-time route optimization in Kampala, Uganda, integrating geospatial traffic data from the Google Maps Directions API with a Random Forest Classifier (RFC) deployed through a Node.js and Python microservice architecture. Traffic congestion is quantified using a ratio-based congestion score derived from baseline and traffic-adjusted travel durations. Model evaluation on 421 route samples yields an overall classification accuracy of 78%, with a Receiver Operating Characteristic–Area Under the Curve (ROC–AUC) of 0.87, indicating strong threshold-independent separability, and a precision–recall average precision of 0.92, demonstrating robust positive-class performance under class imbalance. The deployed system achieves near-real-time responsiveness, with average end-to-end response times below two seconds. The results demonstrate that lightweight and interpretable machine learning models can effectively support congestion-aware routing in data- and resource-constrained urban environments.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Intelligent traffic routing in urban cities using a data-driven machine learning framework for real-time route optimization in Kampala, Uganda

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月19日

查看全文
  • Lillian Tamale,
  • Hedwig Orieba,
  • Swaibu Nzeuliro

摘要

Urban traffic congestion is a persistent challenge in rapidly growing sub-Saharan African cities, where limited sensing infrastructure constrains predictive traffic management. This study presents a data-driven intelligent traffic routing framework for real-time route optimization in Kampala, Uganda, integrating geospatial traffic data from the Google Maps Directions API with a Random Forest Classifier (RFC) deployed through a Node.js and Python microservice architecture. Traffic congestion is quantified using a ratio-based congestion score derived from baseline and traffic-adjusted travel durations. Model evaluation on 421 route samples yields an overall classification accuracy of 78%, with a Receiver Operating Characteristic–Area Under the Curve (ROC–AUC) of 0.87, indicating strong threshold-independent separability, and a precision–recall average precision of 0.92, demonstrating robust positive-class performance under class imbalance. The deployed system achieves near-real-time responsiveness, with average end-to-end response times below two seconds. The results demonstrate that lightweight and interpretable machine learning models can effectively support congestion-aware routing in data- and resource-constrained urban environments.