Using Spatial Autoregressive Lag Model to Predict Crash Rates in Roads Segments (Maximum Likelihood Estimation)

Fatality is a common disorder characterized by car crashes. One of the most significant current discussions in transportation science is to find a precise method to predict car crashes in different areas such as road segments. The past decades has witnessed the rapid development of using statistical methods to predict accidents. These rapid changes  have a serious effect on reduction of  the number of fatalities in car crashes. However, far too little attention has been paid to the role of spatial dependence between roads segments. This paper compares spatial autoregressive methods for considering spatial dependence in accident rate prediction. This paper has been divided into 4 parts. The first part deals with collecting accident data along the road. The second part focused on dividing road into the same length segments. Qazvin- Zanjan freeway was considered for this research. The third part deals with accident rates calculation in each segment. The forth part of this study deals with using spatial autoregressive methods to predict the accident rates in each segment and comparison the results with the results of conventional regression models such as linear regression model. The results of this investigation show that spatial dependence consideration lead to improve the accident rate prediction models accuracy.