Abstract: Stock prices are affected by many factors, which poses a great challenge to stock index prediction. In recent years, machine learning has been widely used in the research of stock price prediction. However, the existing methods have some disadvantages such as large relative error and high time complexity. An improved artificial fish swarm algorithm based on gravity search (AFSA_GS) is proposed. AFSA_GS applies the gravity search strategy of calculating mass and acceleration to the visual and step size adjustment of artificial fish respectively, so as to improve the adaptive ability of artificial fish swarm algorithm in the optimization process. AFSA_GS is further used to optimize the relevant parameters of RBF neural network, and the optimized network is used to predict the stock price. Experiments were conducted on the stock data of a number of listed companies. The results show that, compared with the traditional optimization algorithm, AFSA_GS algorithm can be used to optimize RBF neural network, which can obtain better stock prediction performance.

Key words: stock price prediction, artificial fish swarm algorithm, gravitational search algorithm, RBF neural network, visual, step