An Adaptive Neural Network Classifier for Tropical Cyclone Prediction Using a Two-Layer Feature Selector

We are in need of more accurate, automated prediction and classification methods for the determination of weather patterns all over the world, especially for the identification of severe weather patterns such as tropical cyclones (TC). They help to discover hazardous meteorological phenomena, providing an early warning to save people's lives and properties. In this paper, we propose an adaptive neural network classifier to predict the intensity of a tropical cyclone based on associated features, which is preprocessed by a twolayer feature selector. A binary trigger is used to adjust the neural network topology adaptively when necessary by controlling the validity of each hidden node. Experimental results show that our proposed classifier is a preferable one on learning speed and predictive accuracy comparing to other neural algorithms.