Abstract: The quality of teaching assistants’ work is important to students' education and inclusion, so it is of significance to evaluate and improve the performance of teaching assistants. Support vector machines with appropriate parameters may provide good tools for enhancing the recognition accuracy. Some basic knowledge on support vector machines was firstly introduced; then the paper applied the teaching assistant evaluation data set to examine the recognition effects of SVMs with default and chosen parameters, showing that different parameters may produce different evaluation results. Cross validation method and particle swarm optimization were respectively applied to optimize the parameters of support vector machines, both of which enhanced the recognition accuracy. Finally, conclusions and recommendations were given.