133 / 2016-05-09 18:27:05

Stock Price Prediction Based on the Ensemble Neural Networks for B2B Sector of Shenzhen Stock Exchange

10065,6290,5947,10066,10067

Stock market price movement prediction is quite difficult. This is primarily because of stock price time-series data is characterized by uncertainties, nonlinearities, and high-frequency multi-polynomial components. A variety of methods exist for providing stock market price prediction, with artificial neural networks using technical indicators forecasting variables, but the importance of each remains poorly understood. In this paper, we propose a method to improve the quality of stock market price prediction made using Ensemble Neural Networks. This is accomplished by combining technical indicators which are classified into trend indicators, countertrend indicators and other indicators, BP neural networks (BPNN) for designing Ensemble Neural Networks with Principal Component Analysis (PCA) of preprocessing of input data, and fuzzy systems of responses to forecast complex time series. We evaluate capability of the proposed approach by a variety of experiments from the Internet Firms engaged in Business-to-Business (B2B) of Shenzhen Stock Exchange. Simulation results show that the stock market price prediction made by our approach are more accurate than Back-propagation Neural Networks (BPNN).