Learning to forget in an online fuzzy neural network using dynamic forgetting window

Abstract

This proposed architecture of using a Dynamic Window to compute the forgetting factor which would be able to provide thorough analysis of the self-reorganizing approach when applied to time-variant financial market such as S&P-500 index. When handling such large market, drifts and shifts in inevitable and the system require the ability to have self-reorganizing abilities. To increase its accuracy, the proposed architecture uses the variable dynamic window to adjust the forgetting factor accordingly.