Abstract: In order to reveal the causes of different patterns of traffic accidents occurring on plateau mountain roads, and to clarify the key factors affecting different traffic accident patterns and their degree of influence, a Logistic model is constructed based on the statistical data collected from typical highway accidents in plateau mountainous areas. The causes were analyzed by using three traffic accident patterns as dependent variables, namely rear-end collision, head-on collision and scratch, and using seven types of affecting factors as independent variables from three perspectives of traffic environment, road facilities and vehicle types. The results show that there are obvious differences in the causes of different traffic accident patterns: the marking line setting is the protection factor for all of the three patterns, the signal setting is the protection factor for head-on collisions, while the risk factor for rear-end accidents is the motor and non-motor separating belt; what affects most on the rear-end accident is the large passenger car, which is 5.282 times of other vehicles; the medium-sized truck has the greatest impact on the head-on collision, which is five times that of others, but it has the least impact on scratch accidents; the adverse weather affects the rear-end collision 3 to 4 times of normal weather and the probability of a rear-end collision is 1.312 times that of the daytime if there is no street lighting at night.

Key words: traffic safety, cause analysis of accidents, traffic accident patterns, Logistic model, plateau mountain road