Research and application of XGBoost in imbalanced data

As a new and efficient ensemble learning algorithm, XGBoost has been widely applied for its multitudinous advantages, but classification effect in the case of data imbalance is often not ideal. Aiming at this problem, an attempt was made to optimize regularization term XGBoost, algorithm based on mixed sampling proposed. The main idea combine SVM-SMOTE over-sampling EasyEnsemble under-sampling technologies processing, then obtain final model by training ensemble. At same time, optimal parameters are automatically searched adjusted through Bayesian optimization realize prediction. In experimental stage, G-mean area under curve (AUC) values used as evaluation indicators compare analyze performance different methods models. results public set also verify feasibility effectiveness proposed algorithm.