A typhoon tracks classification method based on similarity threshold learning

The classification of typhoon tracks is an important way to analyze the characteristics of typhoons, and it is also an important method to judge the areas and ranges affected by typhoons. The main methods such as subjective identification and K-means clustering are applied to the research of typhoon classification. These methods have some problems such as randomness caused by traditional clustering algorithms, also rely on artificial experience to set model parameters and are not universal due to specific typhoon research. Aiming at the above problems, this paper proposes a typhoon tracks classification method based on similarity threshold learning. Firstly, a hybrid clustering algorithm based on density center of mass is proposed to realize hot-spot mining and reduce the randomness of traditional methods. Secondly, an adaptive motif processing algorithm based on fastDTW algorithms is proposed to realize more intelligent tracks classification. Finally, the typhoon track landing in Zhanjiang is taken as an example to demonstrate and verify the classification effect of this method.