A Novel DEA-XGBoost Hybrid Framework for Superior Bank Branch Efficiency Assessment

Eslami Nosratabadi, Hamid; Belo, Orlando

doi:10.22068/ijiepr.37.4.2711

IN PRESS Back to the articles list | Back to browse issues page

‎ 10.22068/ijiepr.37.4.2711

A Novel DEA-XGBoost Hybrid Framework for Superior Bank Branch Efficiency Assessment

Hamid Eslami Nosratabadi¹

, Orlando Belo²

1- ALGORITMI/LASI Centre, School of Engineering, University of Minho, Braga, Portugal , hamideslami.na@gmail.com
2- ALGORITMI/LASI Centre, School of Engineering, University of Minho, Braga, Portugal

Abstract: (63 Views)

With the growing complexity of banking operations and increasing customer expectations, reliable evaluation of bank branch performance requires methods capable of integrating multidimensional indicators and identifying nonlinear performance patterns. This work proposes a hybrid DEA–machine-learning framework combining Data Envelopment Analysis DEA for relative efficiency assessment with supervised learning for performance classification. The framework is applied to real-world data from 3,124 bank branches of a major commercial bank in Iran, using 30 financial and non-financial key performance indicators. First, an input-oriented CCR DEA model is used to calculate branch-level efficiency scores. These continuous scores are then discretized into four performance classes Efficient, Relatively Efficient, Moderately Efficient, and Inefficient which constitute the target variable for subsequent classification. Several machine-learning models, including XGBoost, Support Vector Machine, Random Forest, Multilayer Perceptron, and a stacked ensemble, are trained and evaluated using an 80/20 train-test split and 10-fold cross-validation. Results show that SVM achieves the highest reported classification accuracy, while XGBoost provides competitive predictive performance together with interpretable feature-importance analysis. The proposed framework therefore integrates DEA-based benchmarking with machine-learning-based classification and diagnostic analysis, providing a practical decision-support approach for identifying performance differences and supporting targeted managerial interventions in bank branch networks.

Keywords: DEA, XGBoost, Bank Branch Performance, Efficiency Evaluation, Hybrid Machine Learning Model, Decision Support System.

Full-Text [PDF 613 kb] (25 Downloads)

Type of Study: Research | Subject: Productivity Improvement
Received: 2026/03/14 | Accepted: 2026/09/12 | Published: 2026/09/12