Predicting Financial Well-Being Using Observable Features and Gradient Boosting

摘要

Financial well-being and its measurement are well researched topics in personal finance, yet there is no universally agreed definition of financial well-being. Machine learning is proliferating into new application domains. In this study we investigate the use of state-of-the-art gradient boosting methods for predicting subjective levels of financial well-being, using the Consumer Finance Protection Bureau (CFPB) National Financial Well-being dataset. To enable interpretability, we identify the most important observable features required for accurate predictions. These important features are then analysed using factor analysis to understand hidden themes in the data.