Fintech predictive modelling using machine learning: An empirical investigation of South Africa

Abstract

A significant change in the financial industry has been brought about by technological advancements and shifting consumer behaviour. The result has been the emergence of fintech. Consequently, this paper constructs predictive models to forecast the adoption of fintech in South Africa. It investigates the predictive strengths of remittances, borrowed funds, age, access to finance, and digital usage on fintech adoption. Using data sourced from Findex and the South African Reserve Bank (SARB), this paper uses machine learning (ML) approaches namely, multi-linear regression and the ElasticNet algorithms to carry out its predictions. The results offer evidence that the selected factors are key drivers of fintech adoption in South Africa. A good fit of the model is checked through the train and test set score while the error is checked through the root mean square error. The prediction capability of the model is reported at 0.99 with a cross validation of 0.97 thus implying that the chosen inputs provide a good explanation for fintech adoption in SA. Fintech firms are therefore urged to incorporate these findings in formulating strategies to attract and retain customers. The findings also contribute towards the operational strategies that can be adopted by fintech firms.