Template-Type: ReDIF-Article 1.0

Author-Name: Christian S. de Leon
Author-Workplace-Name: Bangko Sentral ng Pilipinas, San Beda University-Manila
Author-Email: chansdeleon@gmail.com
Author-Person:

Title: A Meta-Learning Model of Philippine Bank Lending Behaviour

Abstract: This study develops a meta-learning framework to predict the lending behaviour
of Philippine commercial banks using aggregate bank financial ratios and
macroeconomic variables. Five baseline machine learning models - Boosting,
k‑Nearest Neighbours, Neural Networks, Random Forest, and Support Vector
Machine - were employed, with their outputs synthesised through LASSO‑regularised
regression. Results demonstrate that the metamodel consistently achieves superior
accuracy, lower error levels, and closer proximity to perceived bank lending
behaviour. Robustness checks confirmed stability across volatile and low‑variance
regimes using Ridge and Elastic Net, while feature importance highlighted
profitability and asset quality as key drivers of lending behaviour.

Classification-JEL: C45, E58, G21
Keywords: machine learning, metamodeling, bank lending behaviour
Pages: 119-158
Volume: 25
Issue: 2
Year: 2026
File-URL: https://hitelintezetiszemle.mnb.hu/sw/static/file/fer-25-2-st3-de-leon.pdf
File-Format: Application/pdf
Handle: RePEc:mnb:finrev:v:25:y:2026:i:2:p:119-158