چكيده به لاتين
The present study attempt to develop a descriptive Models for prediction of crash severities in urban traffic areas. This method is mainly based on utility Hypothesis, also used binary logit model (logistic regression), and multinomial logit models.
The main assumptions in the selection of effective variables is based on previous studies and available data. Consequently, the traffic variables (volume, speed, and v/c), collision type , surface term, crash characteristic , existence of middle and parking , road category ,geometrical specification are selected to playing role in the intensity of accidents severities.
Pearson correlation test showed no correlation between independent variables and Variance Inflation Factor (VIF) Eventuate multi-collinearity just for variable as vc and volume. Based on results of complete binary model, variables such as etct, pedes, motor, cloud, twoway, S1, length, arterial1, arterial2 and freeway is significance (p-value<0.05). It should be noted, finally the comparison of statistics AIC, BIC and -2LOG L complete model with other models, it's been chosen the final model (limited). multinomial model results show that after payment of the full model among all predictor variables (independent), 12 variable (pedes, motor, etcm, PP, JP, JJ ،JA, arterial2, Riding، VC and Speed) are significant, as a result, will be present in the final model. Among the collision type variables, jj (head to head) have the greatest impact on the dependent variable. The vehicle congestion (VC) reduce the probability of accidents and the severity of the injury and damage.
Keywords: safety, traffic crash severity, discrite choice models, logit models.