Bank Classification Algorithm: Case Study of Ghanaian Banks

Abstract

The assessment of banks performance with respect to their efficiency is normally done using Data Envelopment Analysis (DEA) model. Determining the performance of a bank by using only its overall efficiency without taking into consideration the efficiency at the deposit stage (stage 1) and efficiency at the investment stage (stage 2) can provide misrepresentation results and unfair assessments. Literature on using a two-stage DEA model to measure and classifying the banks using data from Ghana is also scarce. The purpose is to evaluate the efficiency of bank using DEA model and also classify the bank branches into classes using a proposed Bank Classification Algorithm (BC Algorithm). The study adopted a two-stage DEA model to measure the efficiency scores of 444 Ghanaian bank branches. The efficiencies of the banks in collecting deposits as well as investing the deposit were all calculates using the CCR Two-Phase DEA BuildHull algorithm. This algorithm has its package “Robust Data Envelopment Analysis” (rDEA) version 1.2-5 was implemented in R using R studio. The proposed Bank Classification Algorithm (BC Algorithm) was used to classify the banks where efficient bank was designated as bank with efficiency score of 80% or more. For deposit stage, 72.5%of the Banks were in class 2 thus, majority of the banks even though were not efficient in using resources to collect deposit, they were able to achieve overall efficiency. Just about 10.6% of the banks were efficient in collecting deposit from customers and also achieve overall efficiency. For investment stage, 82.9% of the banks were class 2 thus, majority was not efficient deposit (stage 2) but they were able to achieve overall efficiency. Just about 16.7% (74) of the banks were not efficient in investing their deposit and also achieve overall efficiency. Finally, the results of the analysis show that majority of DMUs thus commercial bank branches in Ghana achieve greater percentage (80% or more) of performance/ overall efficiency in their dual role operations. Further work could be carried out by applying a different model or even combining machine learning algorithms such as Decision Tree, Random Forest or Neural Network with the DEA to predict the efficiency scores on these selected banks.