AI, Deep Machine Learning via Neuro-Fuzzy Models: Complexities of International Financial Economics of Crises

This paper addresses one of the five major areas of AI research identified by Domingos in computer science. I demonstrate that certain Artificial Intelligence (AI) and Machine Learning (ML) type of modeling has great relevance for difficult areas of financial economics and complex financial systems analysis. In a neural network and fuzzy set theoretic formal setting, the ML model predicts the currency crises by combining the learning ability of neural networks and the inference mechanism of fuzzy logic. The empirical results show that the proposed neuro fuzzy model can not only provide better prediction (for both insample and out-of-sample data) but also model causally more detailed relationships among the variables through the obtained knowledge base. Additionally, causal structural path analysis can have significant implications for policy making. The (partially identified) causal path scan also be the bases for further theoretical modifications. One interesting feature of this approach is that it points towards the salient causal role of deep inductive learning in the study of financial crises.