Tropical Cyclone Predicting System Model based on Ensemble Learning

Every year, tropical cyclones cause a great deal of devastation and loss of life and property. Traditional strategies for forecasting tropical cyclone tracks pose various problems in terms of forecast reliability and precision as vast quantities of meteorological and tracking data continue to accumulate, Machine learning techniques have recently been competent enough to work with spatial and temporal features from a lot of data while being extremely efficient and accurate for forecasting data in big structures. Ensemble learning is a method of creating and combining several different models to solve a particular computational problem getting the highest precision. An ensemble can perform better and is able to make better predictions with better performance compared to a single model. The existing model uses the Numerical Weather Prediction (NWP) Model with Sanders Barotropic technology, deep learning applied with Convolutional Neural Network, and GRU using a dataset consisting of actual past hurricanes and cyclones occurring in the years 1945 to 2015, where the complex correlation of meteorological variables is not considered. Hence, we proposed, Boosting Tropical Cyclone Prediction Efficiency Through Time Series Analysis using an Ensemble Learning Approach. Hereby enabling support for Auto Correlation Analysis and reducing the dimension of input variables by removing irrelevant noise data. Ensemble Analysis is done efficiently in this use case giving us better prediction accuracy. The proposed system predicts cyclone status using real-world meteorological variables improving the overall accuracy and precision