Title: Financial inclusion dataset classification in Eswatini using support vector machine and logistic regression

Authors: Stephen Gbenga Fashoto; Boluwaji A. Akinnuwesi; Elliot Mbunge; Andile S. Metfula

Addresses: Department of Computer Science, Faculty of Science and Engineering, University of Eswatini, Kwaluseni Campus, Eswatini (formerly Swaziland) ' Department of Computer Science, Faculty of Science and Engineering, University of Eswatini, Kwaluseni Campus, Eswatini (formerly Swaziland) ' Department of Computer Science, Faculty of Science and Engineering, University of Eswatini, Kwaluseni Campus, Eswatini (formerly Swaziland) ' Department of Computer Science, Faculty of Science and Engineering, University of Eswatini, Kwaluseni Campus, Eswatini (formerly Swaziland)

Abstract: Small scale enterprises grow with provision of financial inclusion (FI) schemes for entrepreneurs. This widens their capital base; hence invest more and increase employment rate. We focus on Eswatini FI scheme from 2018; applied SVM and LR to classify FI dataset; discovered degree to which small, micro and medium enterprises (SMMEs) within Eswatini access funds. FI dataset was extracted from Finscope database. We selected parameters; classified FI for Manzini, Hhohho, Lubombo, and Shiselweni in Eswatini using LR with 80% split for training; ten-fold cross-validation. Manzini has ten-fold cross-validation recall rate of 69.4% using SVM and 63.4% using LR; optimal performance of the 80% percentage split recall rate of 73% was for Manzini using SVM and 77.8% using LR. The 80% split outperforms ten-fold cross-validation. Findings reflect that Eswatini Government should pay more attention to enhance FI in Hhohho, Shiselweni and Lubombo and consider mobile money as key indicator for FI.

Keywords: financial inclusion; support vector machine; SVM; logistic regression; confusion matrix; economic governance; small, micro and medium enterprises; SMMEs; Eswatini.

DOI: 10.1504/IJBIS.2023.132811

International Journal of Business Information Systems, 2023 Vol.43 No.4, pp.507 - 527

Received: 25 Jun 2020
Accepted: 09 Aug 2020
Published online: 10 Aug 2023 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article