A Novel Data-driven Approach for Tropical Cyclone Tracks Prediction Based on Granger Causality and GRU

The strong tropical cyclones will make a drastic effect on human life and natural environment. In these days, as meteorological data and monitoring data accumulated to a large amount, traditional methods of predicting tropical cyclones tracks face many challenges about the prediction efficiency and accuracy. Recently, deep learning method has proven to be an efficient and accurate way to forecast time series data. Therefore, this paper proposes a novel data-driven deep learning model based on Granger causality and Gated Recurrent unit(GRU) to predict the tropical cyclones tracks by selecting the meteorological factors that affect the tropical cyclone locations. The model comprises several aspects including data preprocessing, the feature selection layer, and the GRU with a customized batch process. The model was trained using a real-world tropical cyclones dataset from the year 1945 to 2017, and the results proved that our proposed model can improve the prediction accuracy in the experimental scenario.