Research on Typhoon Path Prediction Based on the Wide & Deep Framework
Yuzhuang He, Xingyu Hao, Qiulong Sheng · 2023
The activity path of a typhoon is influenced by various factors, and traditional trajectory prediction methods have difficulty improving the accuracy of typhoon activity path prediction. To address the issue of low accuracy in typhoon path prediction, a heterogeneous fusion model based on the WIDE & DEEP framework, called ST_LSTM and ConvGRU, is proposed to predict the 24-hour typhoon activity path. By fusing two-dimensional and three-dimensional temporal features, the model aims to improve the accuracy of typhoon path prediction. Experimental results show that this method effectively achieves the heterogeneous fusion of temporal features in typhoon path prediction, outperforming other deep learning methods for typhoon path prediction and further enhancing the accuracy of typhoon activity path prediction.