Lili Zhang

Imbalanced Data Problems, Methods, and Applications

Imbalanced data exists across a wide range of business problems, including credit risk valuation, retail customer targeting and segmentation, network threat detection, anti-money laundering, and rare disease diagnosis. Successfully addressing these problems requires identifying and addressing the imbalanced data problem. In this talk, we will review the practical challenges of imbalanced data, discuss some of the foundational, recently published methods to address these challenges and their applications. Specifically, we will discuss our research on novel weighting mechanisms for penalizing the objective function used to train machine learning models on imbalanced data, including recent use cases and practical applications across multiple relevant business problems.