How fuzzy logic affects quant advanced neural networks?
This was a discussion found on Linked In:
Have done only very limited work with fuzzy logic as my philosophical perspective is that the answer is in very complex relationships that can be found though predictive analytics and automated discovery. These relationships are often both precise and subtle. FL to me always seemed more of a tool for use with human generated crisp rules as in the case of creating an expert system.
As always, depending on your aim. I used FL in a concept competing with Neuro in order to demythologize the Z,-Score of Altman. Neuro and Altman are not exceeding 80% (on the same data set Altman even less than 60%) and FL brought me 97% accuracy.
In order to prove its quality the stockpicking in since 2003 real life delivering 130%
as you may find here :