Road Safety Performance Function Analysis With Visual Feature Importance of Deep Neural Nets

doi:10.1109/JAS.2020.1003108

CORC > 自动化研究所 > 中国科学院自动化研究所 > 学术期刊 > IEEE/CAA Journal of Automatica Sinica

	Road Safety Performance Function Analysis With Visual Feature Importance of Deep Neural Nets
	Guangyuan Pan; Liping Fu; Qili Chen; Ming Yu; Matthew Muresan
刊名	IEEE/CAA Journal of Automatica Sinica
	2020
卷号	7 期号:3 页码:735-744
关键词	Deep learning deep neural network (DNN) feature importance road safety performance function
ISSN号	2329-9266
DOI	10.1109/JAS.2020.1003108
英文摘要	Road safety performance function (SPF) analysis using data-driven and nonparametric methods, especially recent developed deep learning approaches, has gained increasing achievements. However, due to the learning mechanisms are hidden in a “black box” in deep learning, traffic features extraction and intelligent importance analysis are still unsolved and hard to generate. This paper focuses on this problem using a deciphered version of deep neural networks (DNN), one of the most popular deep learning models. This approach builds on visualization, feature importance and sensitivity analysis, can evaluate the contributions of input variables on model’s “black box” feature learning process and output decision. Firstly, a visual feature importance (ViFI) method that describes the importance of input features is proposed by adopting diagram and numerical-analysis. Secondly, by observing the change of weights using ViFI on unsupervised training and fine-tuning of DNN, the final contributions of input features are calculated according to importance equations for both steps that we proposed. Sequentially, a case study based on a road SPF analysis is demonstrated, using data collected from a major Canadian highway, Highway 401. The proposed method allows effective deciphering of the model’s inner workings and allows the significant features to be identified and the bad features to be eliminated. Finally, the revised dataset is used in crash modeling and vehicle collision prediction, and the testing result verifies that the deciphered and revised model achieves state-of-the-art performance.
内容类型	期刊论文
源URL	[http://ir.ia.ac.cn/handle/173211/42984]
专题	自动化研究所_学术期刊_IEEE/CAA Journal of Automatica Sinica
推荐引用方式 GB/T 7714	Guangyuan Pan,Liping Fu,Qili Chen,et al. Road Safety Performance Function Analysis With Visual Feature Importance of Deep Neural Nets[J]. IEEE/CAA Journal of Automatica Sinica,2020,7(3):735-744.
APA	Guangyuan Pan,Liping Fu,Qili Chen,Ming Yu,&Matthew Muresan.(2020).Road Safety Performance Function Analysis With Visual Feature Importance of Deep Neural Nets.IEEE/CAA Journal of Automatica Sinica,7(3),735-744.
MLA	Guangyuan Pan,et al."Road Safety Performance Function Analysis With Visual Feature Importance of Deep Neural Nets".IEEE/CAA Journal of Automatica Sinica 7.3(2020):735-744.