Using Supervised Learning Methods to Credit Score Informal Merchants

Access to working capital is a significant challenge for the informal retail sector, where better financial products are not easily accessible. This study aims to address this issue by developing data-driven credit scoring models for informal merchants using supervised learning methods. The study uses data collected from Nomanini, a financial technology company that facilitates loans to informal merchants in Lesotho. The objective of the study is to help Nomanini develop accurate credit scoring models to reduce the risk of lending money to potential defaulters. Logistic regression and support vector machines were used as supervised learning methods to predict the default behavior of merchants. Six (6) logistic regression models and twelve (12) support vector machine models were evaluated based on their default predictive power. The best-performing model was a logistic regression model that used a merchant's credit history as the only feature, resulting in a Gini coefficient of 0.6143. The study's findings can help Nomanini determine the creditworthiness of merchants more accurately and reduce the risk of lending money to defaulters. This, in turn, can help increase access to working capital to support and grow small businesses in the informal sector. In conclusion, the study highlights the potential of using supervised learning methods to develop credit scoring models for informal merchants and contribute towards reducing the financial gap in the informal African economy.

Grantor

School of Economics

Year dc:date.issued

2023

Author dc:creator

• Kazimoto, Derick

Advisor dc:contributor.advisor

• Georg, Co-Pierre

• Economics

Handle dc:identifier.uri

http://hdl.handle.net/11427/39587

OAI identifier oai:identifier

oai:open.uct.ac.za:11427/39587