A Neuro-Fuzzy System (NFS) is an Artificial Neural Network (ANN) that operates on the principles of Fuzzy Logic. One of the key advantages of NFSs is that they can successfully address the problems of explainability and data scarcity besetting traditional ANNs. Thus, it follows that NFSs can enhance the range of Machine Learning applications in finance and the problems deemed inappropriate for the use of traditional Machine Learning tools can be revisited and perhaps solved using NFSs. This paper discusses NFSs and their merits and challenges for financial services.
A recent surge of interest in Machine Learning (ML) tools within the financial services industry has exposed many challenges of ML applications in financial modelling. Two of the most prominent challenges are data scarcity and the lack of interpretability of some ML techniques (e.g., Artificial Neural Networks)  Although a few approaches have been proposed to address these issues, none can yet be regarded as a complete solution. This paper explores an alternative approach with an outstanding record of success in the engineering field: Neuro-Fuzzy Systems, a combination of Artificial Neural Networks and Fuzzy Logic.
Originally postulated by University of California, Berkley Professor Lotfi A. Zadeh in 1965, fuzzy logic has a wide range of applications, from automatic transmissions and washing machines to microwaves and camcorders (see Kosko (1999)). Although it is a relatively new area of science, fuzzy logic is also used in finance to assess credit risk (Baesens et al. (2003)) and asset pricing (Linn and Tay (2007)), among other functions (see Serguieva et al. (2017) for a review). This paper describes Fuzzy Logic in the context of financial modelling and demonstrates how it can resolve the data scarcity and explainability issues that plague many ML applications in the financial sector. Two case examples are provided to illustrate NFSs in practice: credit spread modelling and the well-known “leverage effect” in equity markets.
 See for example: BlackRock shelves unexplainable AI liquidity models (Risk.net, November 12, 2018); Big funds muzzle their AI machines (Risk.net, October 15, 2018);
Machine Learning Dangerous When Data Thin (Risk.net, January 24, 2019).
 Quants call for better grasp of how AI models ‘think’ (Risk.net, July 11, 2018).