Uncovering the Latent Structure of Liquidity Risk Using Machine Learning Clustering Algorithms: Evidence from Vietnamese Commercial Banks

Abstract

This study explores the latent structure of liquidity risk in the Vietnamese commercial banking system through the application of unsupervised machine learning clustering algorithms. Using panel data collected from the financial statements of 31 Vietnamese commercial banks over the period 2009–2024, the study constructs a set of indicators capturing banks’ liquidity characteristics, profitability, and funding structure. Representative clustering techniques, including K-means clustering and Hierarchical clustering, are employed to identify groups of banks exhibiting similar levels of liquidity risk. The empirical findings reveal a relatively clear segmentation within the banking system in terms of liquidity stability, associated with differences in operational scale, business performance, and reliance on short-term funding sources. The applied clustering algorithms effectively highlight the underlying structures embedded in the data, thereby reflecting the heterogeneity of liquidity risk across the system. This study contributes to the growing literature by incorporating a machine learning–based approach into liquidity risk analysis and provides empirical evidence to support supervisory practices and prudential risk management in the banking sector.