A hybrid model based on differential fuzzy logic relationships and imperialist competitive algorithm for stock market forecasting

Abstract

In this study, a new kind of fuzzy set in fuzzy time series' field is introduced. It works as a trend estimator to be appropriate for fuzzy time series forecasting by reconnoitering trend of data appropriately. First, the historical data are fuzzified into differential fuzzy sets, and then differential fuzzy relationships are calculated. Second, differential fuzzy logic groups are established by grouping differential fuzzy relationships. Finally, in the defuzzification step, the forecasts are calculated. However, for increasing the accuracy of the models, an evolutionary algorithm, namely imperialist competitive algorithm is injected, to train the model. A massive stock data from four main stock databases have been selected for model validation. The final project, has shown that outperformed its counterparts in term of accuracy.