In order to represent the road traffic conditions accurately, and overcome the insufficient of single evaluation factor, it is necessary to build traffic congestion evaluation system. Firstly, the average travel speed, the average delay per unit distance, the average stop number per unit distance, the average stops time per unit distance and the saturation of road sections are selected to build traffic congestion evaluation system. The evaluation criteria of five indexes are determined by the regression analysis of floating car data. Then, the principal components analysis method is applied to define the weigh which is based on the monitoring sample information, and the expert scoring method is supplemented to do some amendments. The multi-ingredient assessment is transformed into the single-object decision by the improved fuzzy comprehensive evaluation model, and the traffic congestion index concept is proposed. Finally, the traffic congestion evaluation system is conducted on the district traffic improvement planning.