Combining the advantages of convolutional neural network and long short-term memory neural network, this paper proposes a new vehicle speed prediction algorithm combining CNN, LSTM and attention mechanism on the basis of the driver model. The algorithm consists of two parts: the driver model and the vehicle speed series prediction model, which takes the time series information of the vehicle in the past T seconds as input, firstly identifies the driver’s current driving style through the driver model, and selects the corresponding vehicle speed prediction algorithm for the identified style, and then realizes the prediction of the future M second vehicle speed. Experiments show that the proposed algorithm structure can effectively improve the accuracy of vehicle speed prediction.