Abstract: To reduce the computational complexity of optimal control of linear hybrid switching diffusion systems, a new Monte Carlo statistical prediction method is proposed. Firstly, the original optimal control problem in continuous time is approximated by a Markov decision problem in discrete time using numerical method; secondly, the numerical optimal control policies in some random sub-state spaces are obtained by reflecting boundary technique; and finally, the optimal control policy on the entire state space is predicted by statistical prediction based on the structure of the optimal control policy. The method can decrease the computational complexity and can be extended to cases of high dimension and large state space. Numerical examples illustrate and confirm the effectivity of the proposed method.