Abstract: According to the statistical data of Urumqi City from 2006 to 2010 ,nine different crash types of traffic accidents on urban roadways were selected respectively as the dependent variables .Furthermore ,nine factors were select-ed as the independent variables ,in aspects of road facilities and road environment .Based on Binary Logistic Regression model ,this paper established linear correlative models between crash types and nine affecting factors ,evaluated the model parameters ,analyzed the reliability and fitting degree of the model ,and investigated the impact that different independent variables combination have on the dependent variables .The paper also predicted the risk of each crash types under various conditions by using a multi-Logistic model ,and compared the prediction with the actual cases ,and tested the fitting effi-ciency of the model used .