Tropical Cyclone Intensity and Track Prediction in the Bay of Bengal Using LSTM-CSO Method

Abstract

Tropical cyclones (TC) are extreme weather conditions caused by severe circular storms that originate in warm oceans. They are strong destructive forces that cause disastrous effects on human life and property and lead to economic damage. Therefore, it is necessary to forecast the TC intensity to avoid the issues. This study proposes a TC intensity forecast using Long-Short Term Memory (LSTM) with Cat Swarm Optimization (CSO). The LSTM method was optimized using the Cat Swarm Optimization technique to improve accuracy and reduce prediction errors. In this study, the prediction was carried out using the latitude, longitude, pressure, and wind speed of tropical cyclones from 2003 to 2019 in the Bay of Bengal. The performance of the proposed system was evaluated using the performance metrics, such as accuracy, Root Mean Square Error (RMSE), Average Absolute Position Error, Mean Absolute Error (MAE), and Area Under Receiver Operating Characteristic Curve (AUROC). The performance of the proposed system is compared with the results of other traditional methods, and the results show that the LSTM-CSO method outperforms other methods in TC intensity and track prediction.

Keywords

• Tropical cyclone
• hyper-parameter tuning
• cat swarm optimization
• long-short term memory
• pressure
• wind speed