A New Approach in the Examination on the Day-of-the Week Effect Using Fuzzy Rule Based Regressions

In this paper a new approach for the investigation of the week effect is proposed. Specifically, fuzzy dummy variables are created based on three fuzzy membership functions. The first one is the log-sigmoid, the second is the Gaussian and the last one is the triangle fuzzy membership function of the stock returns of each trading day of the week. The results are compared with those of the asymmetric GARCH models. The main advantage of this approach is to capture the Monday and the reverse Monday effect or any other daily pattern based on specific rules and intervals of stock returns. Furthermore the fuzzy regressions are not determined by the disturbance term and we show that there are not autocorrelation and heteroskedasticity problems as also ARCH effects with simple ordinary least squares method. This methodology is very useful for financial traders and fund managers to allocate their assets on the portfolio by buying and selling when a specific rules takes place, in an optimum way and with purpose the maximization of the profits