Scholarlink

<p>The prediction of the severity of traffic accidents is an important part of road safety management, especially in regions where the nature is heterogeneous. The proposed research introduces a hybrid machine learning model, which combines a K-means clustering model with Artificial Neural Networks (ANN) to evaluate the classification of data through the subdividing it into more homogeneous clusters.The 4,535 Baghdadi accidents dataset was divided into three groups, and ANN models were trained independently on each cluster. Findings indicate that the hybrid method was not a consistent performer in comparison to the base ANN. However, an obvious improvement was occuerd on cluster 2, where the accuracy increased to 0.70 as compared to the baseline accuracy of 0.596. On the contrary, the performance of the cluster was about 0.55; meanwhile, cluster 1 showed no meaningful improvement (0.59). The results indicate that clustering provides ANN with an improved performance when the resulting clusters are homogenous in the internal structure, suggesting that hybrid modelling could be selectively used in regard to the specifics of the data. This study provides practical insights for developing targeted, data-driven strategies in traffic safety analysis.</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 Machine Learning Framework Using K-means Clustering and Artificial Neural Networks for Traffic Accident Severity Classification

  • 期刊论文

  • 发表日期：

    2026年4月6日

查看全文
  • Hameed A. Mohammed,
  • Ali Jamal Mahdi,
  • Ali Ahmed Mohammed

摘要

The prediction of the severity of traffic accidents is an important part of road safety management, especially in regions where the nature is heterogeneous. The proposed research introduces a hybrid machine learning model, which combines a K-means clustering model with Artificial Neural Networks (ANN) to evaluate the classification of data through the subdividing it into more homogeneous clusters.The 4,535 Baghdadi accidents dataset was divided into three groups, and ANN models were trained independently on each cluster. Findings indicate that the hybrid method was not a consistent performer in comparison to the base ANN. However, an obvious improvement was occuerd on cluster 2, where the accuracy increased to 0.70 as compared to the baseline accuracy of 0.596. On the contrary, the performance of the cluster was about 0.55; meanwhile, cluster 1 showed no meaningful improvement (0.59). The results indicate that clustering provides ANN with an improved performance when the resulting clusters are homogenous in the internal structure, suggesting that hybrid modelling could be selectively used in regard to the specifics of the data. This study provides practical insights for developing targeted, data-driven strategies in traffic safety analysis.