Stock Trends Forecasting by Multi-layer Stochastic ANN Bagging

Abstract

Stock market prediction is an attractive topic for researchers and inventors. Many researchers use machine learning, especially the artificial neural network, to forecast the trends of the stock market. And the steps to obtain better prediction never stop. In this paper, we propose a novel model, multi-layer stochastic ANN bagging, by integrating bagging and artificial neural network. The stock data includes strong noise. We use weak artificial neural networks to get some information without over-fitting and get better results by combining the weak results together using optimized bagging. Our model can achieve three percent to fifteen percent improvements on Standard and Poor's five hundred index prediction and other index prediction compared with the support vector machine, artificial neural network, artificial neural network model optimized by genetic algorithms and Random Forest. Our experiment results show that our model is a promising alternative to the stock market prediction.