Scholarlink

<p>As the number of road accidents in India is on the rise, intelligent mechanisms that can forecast the risk of accidents and help in making decisions based on the data are urgent. This study hypothesises a hybrid approach that integrates Multi-Criteria Decision-Making (MCDM) with TOPSIS method and a Gated Recurrent Unit (GRU)-based deep learning model on time-series prediction of the risk of road accidents. The system receives traffic accident data via a user-friendly web interface created with Streamlit, data is preprocessed by the system, risk scores are calculated with MCDM, and predictions of future risk scores of accidents are made with GRU. The model is trained and tested using real-world traffic data of the Indian traffic and it shows efficient learning with low RMSE and consistent predictive results. The web implementation also renders it more usable and understandable to the policymakers and traffic departments. The experimentation has shown that our method is a sound one and the tool can be scaled by applying it to smart city environments. The proposed hybrid model combines TOPSIS-based MCDM and GRU deep learning to predict road accident risks using time-series Indian traffic data. A Streamlit web interface enables near real-time CSV uploads, risk score visualization, and forecasting, improving accessibility for policymakers and smart city planners. The model achieved a low RMSE (0.0917) and outperformed baseline methods, validating its capability for accurate, interpretable, and scalable accident risk forecasting. The proposed hybrid framework improves prediction accuracy by approximately 7.4% compared to the baseline GRU model while maintaining computational efficiency.</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 MCDM–GRU framework for Indian road accident risk prediction with real-time web deployment

  • 期刊论文

  • 发表日期：

    2026年7月20日

查看全文
  • Mayura M. Yeole,
  • Anand Kudoli,
  • Rahul Patil,
  • Mahesh Sonawane

摘要

As the number of road accidents in India is on the rise, intelligent mechanisms that can forecast the risk of accidents and help in making decisions based on the data are urgent. This study hypothesises a hybrid approach that integrates Multi-Criteria Decision-Making (MCDM) with TOPSIS method and a Gated Recurrent Unit (GRU)-based deep learning model on time-series prediction of the risk of road accidents. The system receives traffic accident data via a user-friendly web interface created with Streamlit, data is preprocessed by the system, risk scores are calculated with MCDM, and predictions of future risk scores of accidents are made with GRU. The model is trained and tested using real-world traffic data of the Indian traffic and it shows efficient learning with low RMSE and consistent predictive results. The web implementation also renders it more usable and understandable to the policymakers and traffic departments. The experimentation has shown that our method is a sound one and the tool can be scaled by applying it to smart city environments. The proposed hybrid model combines TOPSIS-based MCDM and GRU deep learning to predict road accident risks using time-series Indian traffic data. A Streamlit web interface enables near real-time CSV uploads, risk score visualization, and forecasting, improving accessibility for policymakers and smart city planners. The model achieved a low RMSE (0.0917) and outperformed baseline methods, validating its capability for accurate, interpretable, and scalable accident risk forecasting. The proposed hybrid framework improves prediction accuracy by approximately 7.4% compared to the baseline GRU model while maintaining computational efficiency.