股票市场的集成预测方法研究

Research on Integrated Forecasting Method of Stock Market

摘要: 为了研究集中单一预测方法和组合预测方法的预测的效果。我们分别选取逻辑回归模型、决策树模型、多元神经网络三种单一的预测方法进行预测。为了提高金融股票预测价格的稳定性与有效性，我们把单一方法进行集成，采用随机森林和循环神经网络两种集合方法进行测定。结果表明集成预测的效果稍优于单一预测，而选取较好的单一方法进行集合预测会达到一个更好的效果，基于大数据的循环神经网络具有良好的自适应性和很强的学习能力，在对比的集中方法中是最有效的。

Abstract: In order to study the forecasting effect of centralized single forecasting method and combined forecasting method. We select three single forecasting methods, logistic regression model, decision tree model and multivariate neural network, respectively. In order to improve the stability and effectiveness of financial stock price prediction, we integrate the single method and use random forest and recurrent neural network to measure. The results show that the effect of ensemble prediction is slightly better than that of single prediction, and the selection of a better single method for ensemble prediction will achieve a better effect. The circular neural network based on big data has good adaptability and strong learning ability, and is the most effective among the comparative centralized methods.