The Development of a Consensus Machine Learning Model for Hurricane Rapid Intensity (RI) Forecasts with Hurricane Weather Research and Forecast (HWRF) Data

This study mainly focuses on developing a Consensus Machine Learning (CML) model based on six classical ML algorithms for hurricane intensity change forecasts. CML was designed for generating probabilistic predictions for four intensity change categories: Weakening, Neutral, Intensifying, and Rapid Intensification (RI). The selected ML algorithms for CML development were Logistic Regression, k-Nearest Neighbors, Decision Trees, Random Forest, Support Vector Machine, and Artificial Neural Networks. The analysis data and the six-hour forecast data from the Hurricane Weather Research and Forecasting (HWRF) model were used as the raw data for developing the CML. In this project, 20 attributes from the analysis data and 34 attributes from the forecast data were extracted from the multidimensional HWRF raw data. The attribute sets were rebalanced with Edited Nearest Neighbor-Synthetic Minority Oversampling Technique (ENN-SMOTE) before being used for training, validation, and testing. Our analysis shows that the CML model, the combination of the six ML models, can increase the predictability for all intensity change categories.The CML model can reach about 47% recall rate but with less than 50% false alarm ratio, which outperformed the current operational RI models. Therefore, the CML model has potential to be an operational model with the proper training dataset containing RI-relevant attributes.