基于随机森林的极限学习机集成学习算法

The algorithm first performs multiple random sampling on the telemetry data to construct multiple training subsets. Then, multiple feature samplings are applied to each training subset, and corresponding sub-classifiers are trained. Based on the input features of the optimal sub-classifier within each subset, the probabilities and weights of the feature sampling are updated. Finally, the validation scores of all sub-classifiers are sorted, and a certain proportion of RFELM (Random Forest-based Extreme Learning Machine) are selected to form a classifier ensemble. The ensemble predicts the labels of the data using weighted voting.