The prediction of stock value by using the proposed fuzzy neural network and hybrid algorithm

The prediction of stocks on the stock market and how the symbols are changed are one of the most applied and popular researches. By predicting the symbols with the least error, a person can succeed in the stock market. In this paper, a new network including a neural-fuzzy sink function and an improved grasshopper optimization algorithm was used to predict the value of symbols. In this regard, to predict and model the stock symbols, black-box modeling and AR (Autoregressive) model were used. Model order was determined by using the gray wolf algorithm. To optimize the network’s linear parameters, a hybrid algorithm comprising of least square algorithm for initialization, recursive least square for online training, and a grasshopper optimization algorithm was used to optimize nonlinear parameters. The simulation illustrated that by providing a new structure, the gray wolf algorithm can determine the order of the model and the terms with the most impact on the steel symbol, effectively. In addition, the proposed network and algorithm had less error than other methods such as neural networks for predicting stock value, and the grasshopper optimization algorithm converged with the adaptive learning rate more rapidly.