Application of Explainable Artificial Intelligence technique to model the predictors of South African SMMEs resilient performance during the Covid-19 pandemic

Various studies have been carried out to establish the key drivers impacting small enterprise sustainable performance in developing countries. Despite many policy-oriented studies to uncover the factors influencing SME resilience in emerging markets, these firms continue to register high failure rate, which has been further exacerbated by the Covid-19 pandemic. Guided by a history of linear- and log-linear econometric model estimation that ignores potential network effects, our study extends the literature by implicating SMME resilience as a production network. Utilising data from both incubated and non-incubated SMMEs, marking a departure from traditional linear econometric models, radial basis function artificial neural network algorithm was invoked to establish the drivers of SMME resilience during Covid-19 regime. The study extends the literature by implicating eXplainable Artificial Intelligence (XAI) methods. Specifically, optimal SHapley Additive Explanations values (SHAP values) were computed to enhance the prediction output from the machine learning algorithm. The XAI analytics provide insightful findings on the key drivers which influenced the resilience of SMMEs during the Covid-19 pandemic. The importance of innovation through introduction of new products, company age and higher number of marketing mediums is confirmed however total assets, analytics, educational level and number of workers surfaced as a threat to these enterprises’ sustainable performance. The study recommends that both the government and SMEs should leverage XAI to identify their heterogeneous attributes and inform intelligent decision-making which necessities their resilient performance. Key Words:Artificial Neural Networks, Covid-19, Explainable Artificial Intelligence, SMME Resilience, SHAP Values