Scholarlink

<p>Residential traffic safety poses critical challenges due to dense pedestrian activity, irregular road layouts, and dynamic vehicle movement. Traditional static traffic control measures are often inadequate for such complex environments. This study presents a hybrid expert system that integrates Artificial Neural Networks (ANN), Fuzzy Neural Networks (FNN), and rule-based reasoning to predict and prevent traffic accidents in residential areas. The system utilises multi-source data—including road geometry, traffic density, and weather conditions— to generate short-term (real-time–24 h)<b>,</b> medium-term (1–14 days), and long-term (15–90 days) accident-risk forecasts. Unlike conventional time-series models, the system leverages a hybrid data strategy, combining historical accident records, simulation-based scenarios, and real-time environmental inputs, to provide robust forecasts without requiring sequential tracking of prior risk scores<b>.</b> Applied to real-world data from Baghdad, the model achieved strong predictive performance, with AUC scores of 0.84, 0.87, and 0.78 across the respective time horizons. Beyond prediction, the system supports proactive interventions, such as optimised signal timing and targeted pedestrian safety enhancements, validated through traffic simulations. The results demonstrate the system's potential to inform intelligent, data-driven planning for accident prevention in urban neighbourhoods.</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

A Hybrid Expert System for Predicting and Controlling Traffic Crashes in Residential Areas

  • 期刊论文

  • 发表日期：

    2026年1月27日

查看全文
  • Ali Ahmed Mohammed,
  • Yousif Al Mashhadany,
  • Ali Amer Ahmed Alrawi,
  • Sameer Algburi,
  • Ihab Mahmood Abdulhadi,
  • Hussin A. M. Yahia,
  • Taleb Eissa

摘要

Residential traffic safety poses critical challenges due to dense pedestrian activity, irregular road layouts, and dynamic vehicle movement. Traditional static traffic control measures are often inadequate for such complex environments. This study presents a hybrid expert system that integrates Artificial Neural Networks (ANN), Fuzzy Neural Networks (FNN), and rule-based reasoning to predict and prevent traffic accidents in residential areas. The system utilises multi-source data—including road geometry, traffic density, and weather conditions— to generate short-term (real-time–24 h), medium-term (1–14 days), and long-term (15–90 days) accident-risk forecasts. Unlike conventional time-series models, the system leverages a hybrid data strategy, combining historical accident records, simulation-based scenarios, and real-time environmental inputs, to provide robust forecasts without requiring sequential tracking of prior risk scores. Applied to real-world data from Baghdad, the model achieved strong predictive performance, with AUC scores of 0.84, 0.87, and 0.78 across the respective time horizons. Beyond prediction, the system supports proactive interventions, such as optimised signal timing and targeted pedestrian safety enhancements, validated through traffic simulations. The results demonstrate the system's potential to inform intelligent, data-driven planning for accident prevention in urban neighbourhoods.