Safety modelling and approaches to accident reduction on rural roads, case study Kerman Province Roghayeh Abdollahi Date of Submition 2015.09.05 Department of Traortation Engineering Isfahan University of Technology, Isfahan 84156-83111, Iran Degree: Master of Science language: Persian Supervisor: Dr Hossein Haghshenas, H_haghshenas@cc.iut.ac.ir Advisor: Dr Reyhaneh Rikhtehgaran, R_rikhtehgaran@cc.iut.ac.ir Iran's accidents fatality rate is 35.8 per hundred thousand inhabitants in a way that traffic accidents in Iran are introduced as the third cause of death in the amount of almost 5 times the world average. The first step to reduce road accidents is to know factors affecting accidents. This recognition is possible using analyses and in a more advanced state, with modeling variables affecting accidents. This study aims is identifying the factors that cause accidents with an emphasis on the role of modeling accidents. In this regard, accident models and correlation test are discussed for the two categories of information. The first category is related to general information of all roadand the second category is related to local information of roadand geometric design themes. In the first category of modeling the relations in Highway Safety Manual (HSM) has been validated for Iran. Then with different modelings, including taking iiration from the relations of these manuals, the linear regression models with the integration of main roads and highways with the help of auxiliary variable, and using models with entering the traffic variables, Factorizing Correlated Variables and applying generalized linear regression models (GLM), modeling process is improved. Finally, the most appropriate model for these data has been selected as corrective-integrated linear regression model with importing traffic data and applying the highways and the main road information at the same time with taking into account the auxiliary variable. In this model, average daily traffic in the year, length of traffic axes, the percentage of unauthorized distance in main roads and highways and the percentage of unauthorized speeding on the highways as the most effective variables are involved in accidents. Then, accident hot-spot have been identified according to two indices of accidents frequency and the Compensation Equivalent Index. In the second category of modeling, to study the geometric factors contributing to accidents, geometric design of these roads such as horizontal and vertical alignment characteristics, topography of the region type, number of accesses, etc., in pieces with a length of 5 km of these roads, collecting and/or modeling data with linear and generalized linear models, the most appropriate model is selected. Poisson’s generalized linear model with operating parameters of vertical alignment characteristics, with variables of number of vertical alignment and the average difference of the slope of the vertical alignment and mean difference of the total slope and the factor of region type, with the variables of existence of mountain and the average longitudinal slope of land and foothill of region according to the average longitudinal slope is selected as the most acceptable model. To reduce the occurrence of accidents in the final chapter of this study, guidelines and recommendations have been made in accordance with models and analyses. Keywords : Safety Engineering, Linear Regression, Nonlinear Regression Model, HSM Model, Generalized Linear Regression (GLM), Factor analysis.