To address the problem of nonlinear distortion of high-speed transmission in the visible light communication system, a deep learning based nonlinear distortion compensation method of high-speed transmission in the visible light communication system is proposed.Firstly, a few forward and backward neighbor symbols of the target symbol in the high-speed frames are extracted by a sliding window, which construct a context symbol sequence including the target symbol.Then, the one-dimensional convolutional neural network is used to extract the space feature of the context sequence, and the bi-directional long short term memory network is used to extract the forward and backward temporal dependency. Experimental results show that, the proposed method compensates the distortion of the visible light signals, and it reduces the bit error rate of high-speed transmission in the visible light communication system.