ADASYN-LOF Algorithm for Imbalanced Tornado Samples

Early warning and forecasting of tornadoes began to combine artificial intelligence (AI) machine learning (ML) algorithms improve identification efficiency in the past few years. Applying detect usually encounters class imbalance problems because are rare events weather processes. The ADASYN-LOF algorithm (ALA) was proposed solve problem tornado sample sets based on radar data. adaptive synthetic (ADASYN) sampling is used by increasing number minority samples, combined with local outlier factor (LOF) denoise samples. performance ALA tested using supporting vector (SVM), neural network (ANN), random forest (RF) models. results show that can noise immunity models, significantly increase recognition rate, have potential early time. more effective preprocessing imbalanced data SVM ANN, compared ADASYN, Synthetic Minority Oversampling Technique (SMOTE), SMOTE-LOF algorithms.