An improved hybrid algorithm based on PSO and BP for stock price forecasting

Abstract

Stock price prediction is the main concern for financial firms and private investors. In this paper, we proposed a hybrid BP neural network combining adaptive PSO algorithm (HBP-PSO) to predict the stock price. HBP-PSO takes full use of the global searching capability of PSO and the local searching advantages of BP Neural Network. The PSO algorithm is applied for training the connection weights and thresholds of BP, in order to take advantage of BP, each particle in PSO swarm will be optimized by error correcting method of BP in probability. The trained BP neural network is used to predict the stock price. The empirical analysis using the real data of Chinese stock market demonstrates the feasibility and effectiveness of this method.