دانلود رایگان مقاله لاتین مدلسازی فرکانس تصادف توسط شبکه عصبی از سایت الزویر


عنوان فارسی مقاله:

قانون استخراج از شبکه عصبی بهینه سازی شده برای مدل سازی فرکانس تصادف وسایل نقلیه


عنوان انگلیسی مقاله:

Rule extraction from an optimized neural network for traffic crash frequency modeling


سال انتشار : 2016



برای دانلود رایگان مقاله مدلسازی فرکانس تصادف توسط شبکه عصبی اینجا کلیک نمایید.





مقدمه انگلیسی مقاله:

1. Introduction

In recent decades, numerous models of crash frequency have been proposed to model the relationship between crash frequency at road segments or intersections and risk factors related to traf- fic and geometrical characteristics of the sites. Most studies of this sort have employed statistical count modeling techniques, since these models provide explicit forms for the random, discrete and non-negative nature of counting crash data and the effects of major contributing factors on crash occurrence. In addition to statistical models, some artificial intelligence (AI) models have been proposed (Chang, 2005; Li et al., 2008). As a common class of AI models, neural network (NN) models have been successfully used in many fields of transportation research, including highway safety analysis (Karlaftis and Vlahogianni, 2011).In modeling crash frequency, NNs are able to approximate the potential nonlinear and complicated relationship between crash frequency and risk factors. Several studies have demonstrated the better model fitting and predictive performance of NNs over traditional negative binomial (NB) models, in which nonlinear safety effects of risk factors have been identified (Chang, 2005; Xie et al., 2007). The recently-developed random parameters (Anastasopoulos and Mannering, 2009) and Markov switching (Malyshkina et al., 2009) count models indicate that loosening the constraint of fixed parameters could significantly improve their performance on modeling crash frequency, which also partially reflects the existence of nonlinear relationship in crash modeling. However, NNs have two primary drawbacks that limit their application to traffic safety research, including the so-called “blackbox” characteristic and the possible over-fitting problem (Xie et al., 2007). The black-box characteristic has limited NNs’ ability to explicitly illustrate the effects of explanatory variables on crash frequency. Even for studies using sensitivity analysis, the impacts on safety of each risk factor cannot be systematically or globally interpreted either. To overcome this problem, a more general approachis to extract the knowledge from the NNs. Using regression analysis, Setiono and Thong (2004) proposed a rule extraction method that generated a group of piecewise linear functions to approximate NNs. This method may be adopted in road safety analysis to clarify the relationship between network output(s) and input risk factors.



برای دانلود رایگان مقاله مدلسازی فرکانس تصادف توسط شبکه عصبی اینجا کلیک نمایید.






کلمات کلیدی:

[PDF]Fig. 2 iranarze.ir/wp-content/uploads/2016/09/E3.pdf Rule extraction from an optimized neural network for traffic crash frequency modeling. Qiang Zenga,b, Helai Huangb,∗, Xin Peic, S.C. Wongd, Mingyun Gaoe. [PDF]The Statistical Analysis of Crash-Frequency Data - CiteSeerX citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.471.410&rep=rep1...pdf by D Lord - ‎2010 - ‎Cited by 670 - ‎Related articles researchers have sought to develop separate crash-frequency models for .... becomes smaller as traffic flow increases (Maher et al., 1993; Lord 2002; Lord et al., ..... whereas for Bayesian neural networks, the weights follow a probability ... Rule extraction from an optimized neural network for traffic crash ... hub.hku.hk/handle/10722/237009 Dec 20, 2016 - Title, Rule extraction from an optimized neural network for traffic crash frequency modeling. Authors ... Accident Analysis & Prevention, 2016, v. [PDF]An Artificial Neural Network Model for Road Accident Prediction: A ... https://www.uni-obuda.hu/journal/Ogwueleka_Misra_Ogwueleka_51.pdf by FN Ogwueleka - ‎2014 - ‎Cited by 6 - ‎Related articles Abstract: Road traffic accidents (RTA) are one of the major root causes of the ... paper, we produce a design of an Artificial Neural Network (ANN) model for ... carried out for the examination and prediction of accidents rate using Nigeria as a. Efficient Transportation and Pavement Systems: Characterization, ... https://books.google.com/books?isbn=0203881206 Imad L. Al-Qadi, ‎Tarek Sayed, ‎Naser Alnuaimi - 2008 - ‎Technology & Engineering First by predicting traffic crashes on the freeway mainline using on-line loop detector ... The results of the predictive models were found to be consistent with the ... at estimation of crash frequency or rate on freeway sections through aggregate ... methodologies have been explored e.g., Probabilistic neural networks (PNN) ...