| 摘 要: 近年来,由交通事故引起的人身伤害和经济损失比例逐渐上升,准确预测交通事故的严重程度对于交通安全的管理和控制至关重要。文章创新性地提出了一种具备可解释性的深度神经模糊系统(Deep Neural Fuzzy System,DNFS)对交通事故的严重程度进行预测。该系统通过学习历史数据中的特征相关性和模糊规则,实现对输入特征的降维,使其能更准确地捕捉特征之间的潜在关系。实验结果表明,DNFS对轻微事故、严重事故和致命事故的预测准确率分别达到91%、93%和93%,在多类别预测任务中展现出卓越的性能。 |
| 关键词: 交通事故严重程度;深度学习;可解释性 |
|
中图分类号: TP311
文献标识码: A
|
|
| Research on predicting the Traffic Accident Severity Based on Deep Neural Fuzzy System |
|
WANG Yuanyuan, SHI Donghui, GAN Shuling
|
(School of Electronic and In f ormation Engineering, Anhui Jianzhu University, Hefei 230601, China)
katery1010@163.com; donghui_shi@163.com; 707668439@qq.com
|
| Abstract: In recent years, the proportion of personal injuries and economic losses caused by traffic accidents has been gradually increasing. Accurately predicting the severity of traffic accidents is crucial for the management and control of traffic safety. This paper innovatively proposes a Deep Neural Fuzzy System (DNFS) with interpretability to predict the severity of traffic accidents. By learning the feature correlations and fuzzy rules in historical data, the system achieves dimensionality reduction of input features, enabling it to more accurately capture the underlying relationships between features. Experimental results show that DNFS achieves prediction accuracies of 91% , 93% , and 93% for minor accidents, serious accidents, and fatal accidents, respectively, demonstrating outstanding performance in multi-class prediction tasks. |
| Keywords: traffic accident severity; deep learning; interpretability |
