A Hybrid Markov Decision Process (MDP) Model for Predicting Liquidity Profiles of US Regional Banks
Open Access DepositedThis praxis addresses the critical issue of accurately predicting liquidity for US regional banks during refinancing cycles. Regional banks are a vital part of the local economy, and they suffer more risk than larger financial institutions primarily because they serve a regional market, are less diversified structurally, and borrow from a limited number of potential sources. Inherent in this environment is increased risk exposure as they have limited access to capital should the local economy decline. Present liquidity prediction methodologies failing to consider the complexity of the nonlinear, spatial and temporal characteristics of liquidity profiles, leaving regional banks highly vulnerable to misjudged liquidity needs and refinancing costs, and financial instability.This research proposes to reduce the predictive void by presenting a hybrid prediction framework based on Markov Decision Processes (MDP) and advanced machine learning algorithms in Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Long Short-Term Memory (LSTM). This framework leverages the structured decision-making capabilities of MDP and the robust predictive power of machine learning to enhance accuracy and timeliness in liquidity forecasts. The model is validated using liquidity profiles of regional banks listed in the S&P Regional Banks Select Industry Index (KRE) from 2013 to 2023, focusing on key macroeconomic variables such as interest rate spreads, credit spreads, and bond yield volatility. The hybrid framework couples machine learning accuracy with the MDP’s dynamic decision optimization. Its advantage lies in policy performance, not raw predictive lift, allowing earlier intervention and improved liquidity resilience during refinancing cycles.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.