Imbalance-XGBoost: Leveraging Weighted and Focal Losses for Binary Label-Imbalanced Classification with XGBoost

Abstract The paper presents Imbalance-XGBoost, a Python package that combines the powerful XGBoost software with weighted and focal losses to tackle binary label-imbalanced classification tasks. Though a small-scale program in terms of size, the package is, to the best of our knowledge, the first of its kind which provides an integrated implementation for the two loss functions on XGBoost and brings a general-purpose extension to XGBoost for label-imbalanced scenarios. In this paper, the design and usage of the package are discussed and illustrated with examples. Furthermore, as the first- and second-order derivatives of the loss functions are essential for the implementations, the algebraic derivation is discussed and it can be deemed as a separate contribution. The performances of the methods implemented in the package are extensively evaluated on Parkinson’s disease classification dataset, and multiple competitive performances are presented with the ROC and Precision-Recall (PR) curves. To further assert the superiority of the methods, the performances on four other benchmark datasets from the UCI machine learning repository are additionally reported. Given the scalable nature of XGBoost, the package has great potentials to be broadly applied to real-life binary classification tasks, which are usually of large-scale and label-imbalanced.