Machine_Learning_tor_int

Contains example code from the binary classification machine learning methods used in research to explore tornado intensity prediction approaches using radar data and the near-storm environment of ongoing thunderstorms. A detailed example of a machine learning pipeline using Random Forests and nested cross-validation is demonstrated as well as its utility for exploring machine learning with smaller datasets given the limiting factors of certain observational datasets such as from radar. The pipeline includes cross-validated recursive feature elimination, model training, calibration/model reliability exploration, and hyper-parameter tuning in an inner-cv fold, and model evaluation on testing data in the outer fold which includes several binary classification metrics and diagrams and a detailed exploration of feature importance and model interpretability.