基于EMD和ANFIS的股票市场短期预测

结合经验模态分解（empiricalmodedecomposition，EMD）算法和自适应神经模糊推理系统（adap-tiveneuralfuzzyinferencesystem，ANFIS）算法应用于股票市场预测,提出了一种新的股票市场的预测模型,即EMD-ANFIS的多步预测模型。首先应用EMD算法把原始数据分解成不同尺度的基本模态函数（IMF）和残差（RES）,然后通过ANFIS算法对生成的各个IMF和RES进行自适应神经模糊推理,再把各个预测结果进行简单的聚合作为股票的预测价格,并与传统的预测方法进行比较,实验证明了EMD-ANFIS的多步预测模型具有更高的预测精度。

Short-term Stock Market Prediction Based on EMD and ANFIS

By Integrating Empirical Model Decomposition （EMD） with Adaptive Neural Fuzzy Inference System （ANFIS） theory, a new stock market forecasting model, namely EMD-ANFIS Multi-step Model is put forward. Firstly, EMD theory is employed to decompose original stock market time series into residual （RES） and Intrinsic Model Functions （IMF） of different scales. Then ANFIS theory is employed for the adaptive neuro-fuzzy inference processing of IMF and RES. Finally, by simple addition, the sum of the predicted results is considered as the predicted stock price. Compared with traditional forecasting methods, this experiment proves that EMD-ANFIS method can effectively improve the prediction accuracy.