The analysis, modeling and prediction of historical flight path data of air vehicles are conducive to understanding the regional organization characteristics of air traffic, grasping the air situation in advance and formulating corresponding measures in time. In this paper, a segmented clustering algorithm model based on improved D-P algorithm is proposed to solve the problems of track detail information loss, poor clustering effect and low universality of the previous clustering model. In the track prediction model, the temporal convolutional network is used to replace the long-term and short-term memory network to predict the next time position of the track efficiently and accurately. Also, an idea of track quantization approach is proposed to solve the problem of high complexity of the most similar track matching. The experimental results show the effectiveness of the model.