Abstract: Safety and efficiency of highways are impacted by frequent occurrences of dangerous driving behaviors and traffic violations.Based on related data of traffic safety on highways, a graded warning method is developed to evaluate dangerous degree of traffic violations.First, the relationships between traffic accidents and influence factors, including inclement weather, geometric alignments, and types of traffic violation behaviors, are investigated quantitatively.Second, considering weather conditions, traffic conditions, road geometric alignments and traffic violations, a method combining analytic hierarchy process (AHP) with fuzzy comprehensive evaluation is proposed to develop a graded warning model of traffic violations.A case study is further used to validate the proposed method.Results show that the model is effective and can provide supports for initiative traffic warning and management on highways.