Explainable artificial intelligence (XAI) in investment decision-making Golnoosh Babaei* [1] , Paolo Giudici [1]

https://www.academia.edu/academia-ai-and-applications/1/2/10.20935/AcadAI8017 Financial investments are being facilitated by the availability of artificial intelligence, based on advanced machine learning (ML) methods, with high computational power and accuracy. However, despite their high accuracy, ML models do not provide sufficient explanation and, thus, may not be adequate for informed investment decision-making. In this paper, we propose an explainable AI model that can be used to explain investment predictions and, in particular, to predict the expected return of companies, based on their expected profitability and risk. Our proposal is based on Shapley values, which associate each explanatory variable to the variations in predictions that are induced, considering all possible coalitions. To validate our model, we consider real data containing the balance sheets of 2049 small and medium enterprises (SME). The analysis of the data reveals that the expected return of the SMEs can be effectively predicted and explained by a set of financial indicators deduced from their balance sheets.