Research | Zhang Hongzhe: Latent Similarity-Enhanced Credit Risk Prediction

Release time：09 February 2026

In the current era of rapid expansion in the consumer credit market, accurately predicting default risk is a core challenge for financial institutions. Traditional prediction models are often limited to known explicit features of users, ignoring the unobserved and hidden intrinsic connections between users, which restricts further improvements in risk control precision.

Against this backdrop, the paper Latent Similarity-Enhanced Credit Risk Prediction, co-authored by Zhang Hongzhe from the School of Management and Economics (SME) at The Chinese University of Hong Kong, Shenzhen (CUHK-Shenzhen), Qian Wei and Fang Xiao from the University of Delaware, innovatively proposes the Latent Similarity-Enhanced (LSE) prediction framework. By integrating observable features with latent similarities, this research characterizes user profiles more comprehensively and demonstrates superior predictive performance and economic benefits compared to mainstream algorithms such as XGBoost and Graph Neural Networks (GNN) in empirical tests. Recently, the study was published in MIS Quarterly, a top-tier international journal in management and economics.