Boundedly rational learning and heterogeneous trading strategies with hybrid neuro-fuzzy models

The present study deals with heterogeneous learning rules in speculative markets
where heuristic strategies reflect the rules-of-thumb of boundedly rational investors. The
major challenge for "chartists" is the development of new models that would enhance
forecasting ability particularly for time series with dynamic time- varying, nonlinear features.
This paper introduces fuzzy learning rules with the incorporation of beliefs, preferences and
idiosyncratic behavioral patterns for decision-making and trading under uncertainty. The
efficiency of a technical trading strategy based on a neurofuzzy model is investigated, in
order to predict the direction of the market for NASDAQ Composite, NIKKEI255 and
FTSE100. Moreover, it is demonstrated that the incorporation of the estimates of the
conditional volatility changes strongly enhances predictability, as it provides valid
information for a potential turning point on the next trading day. The total return of the
proposed volatility-based neurofuzzy model, including transaction costs, is consistently
superior to a markov-switching model, a recurrent neural network as well as to the buy &
hold strategy for all indices. The findings can be justified by invoking either the "volatility
feedback" theory or the existence of portfolio insurance schemes in the equity markets and are also consistent with the view that volatility dependence produces sign dependence.
Overall what leads to optimal prediction is the dynamic update of the expectations and
preferences of the heuristic learning rules combined with the adaptive calibration of the
"degrees-of-belief" that match agent’s "fads".