Abstract: Holding strategies are among the most commonly used operation control strategies.The author presents a new holding strategy.Firstly,the strategy determines a held bus by using a support vector machines (SVM) and Kalman filter-based prediction model.Then,in order to determine the optimal holding time,a model aiming to minimize the user costs is developed and a genetic algorithm (GA) is used to optimize the holding time of a held bus at a stop.Finally the presented holding strategy proposed in this study is illustrated with the microscopic simulation model Paramics,and results show that the dynamic holding strategy is effective compared with traditional holding strategy and no control strategy.