Machine learning for predictive model in entrepreneurship research: predicting entrepreneurial action

This study introduces a method for developing predictive models using machine learning in entrepreneurship research. Machine learning is known to provide a superior performance of prediction by identifying hidden patterns in data through an inductive approach. However, there are very few studies adopting this methodology in social sciences, especially in the field of entrepreneurship. This study investigates the utility of machine learning in entrepreneurship research and proposes a practical method to develop a predictive model using machine learning. For the implementation of this method, as a case study, this study builds a model that predicts entrepreneurial action based on data from the Global Entrepreneurship Monitor (GEM). This study compares the performance of machine learning such as XG boost and artificial neural network (ANN) with traditional statistical method, logistic regression model. Performance indicators such as accuracy, sensitivity, specificity, and area under curve (AUC) were used for evaluation. XG boost showed the highest performance in all indicators except for precision. In the analysis of the variable importance, self-efficacy and opportunity are the most influential factors for predicting entrepreneurial action.