Abstract: The typhoon paths are essentially curves on a two-dimensional plane. According to the similarity of the two typhoon path curves, the numerical similarity and morphology similarity can be judged. Therefore, the dynamic time warping algorithm can be used to select the typhoon path similar to the target typhoon from the historical typhoon database. Meanwhile, considering the time correlation of typhoon path information, we put forword a long short-term memory model. The paper uses the latitude and longitude information of historical typhoon to predict the position information of the typhoon in the next 6 hours. Compared with the traditional method of predicting typhoon path by similarity, the long short-term memory model can effectively improve the prediction accuracy of typhoon paths, and the model is more stable and efficient.

Key words: similarity, DTW, LSTM, trajectories predicting