The use of Firefly Algorithm and Bayesian Regulation technique of optimized Artificial Neural Network to predict stock price in Iran Stock Market

Abstract

Predicting the future stock price has always been considered as an important issue by both buyers and sellers. Hence, Artificial Neural Network (ANN) was used in this study to develop a model pertaining to artificial intelligence in order to predict stock price in Iran Stock Market. Since artificial neural networks should consist of the best network topology to achieve the highest performance, Firefly Algorithm (FA), a meta-heuristic Algorithm, was used to find the optimal structure of network. Finally, Bayesian regulation technique, rather than the conventional teaching techniques, was applied to maintain the more generalized network. In general, Data from three big companies: Iran Khodro Company, Shiraz Petrochemical Company, and Isfahan Steel Company were gathered in span of three years. This paper profited from some parameters, including high price, low price, the opening price, closing price, EMA(5)، EMA(10)، RSI، William R%، Stochastic k%، Stochastic D%، و ROC as network inputs and benefited from the closing stock price in the next days as the neural network as well. After developing a model associated with each company, some parameters such as the root-mean-square error (RMSE), Standard Deviation of error (SD), Absolute average relative deviation (AARD), the regression coefficient (R2) as well as the graphical analysis of relative deviation have been used to examine the accuracy of the developed network. The outcomes of the analysis of the developed neural networks revealed that the mentioned models with great accuracy are able to predict stock price in the subsequent day for the corporations mentioned above.