Based on the traffic data and crash data collected on G15, this paper studied short-term traffic flow risk prediction model on freeways with high proportion of trucks and high proportion of truck crashes. The overall traffic flow parameters, the truck traffic flow parameters and the comprehensive parameters were selected as the risk characteristic variables. The support vector machine was adopted for the modeling and genetic algorithm was used to optimize the parameters. Classification models of different time periods, different risk characteristics variables were got and compared. The results show that the model using the data within 5 to 10 minutes before the accident performs the best. When considering truck factors,the overall prediction accuracy improves 7.1% , the crash rate prediction accuracy improves 6.6% and the false alarm rate is 7.7% lower. Finally, the different importance of characteristic variables was obtained through mean impact value. The results show that truck factors have larger effects on the prediction model. The model in this research can be used to developSearly warningSsystem of traffic security and provide theoretical basis of truck safety management on freeways.