基于误差校正的灰色神经网络股票收益率预测

在股票市场中,准确的股票收益率预测是市场交易各方共同关心的重要问题。由于影响股票市场的因素十分复杂,仅靠建立单一的股票收益率预测模型来提高预测精度是非常困难的。本文对当前股票收益率预测方法存在的不足进行了阐述,并提出了以误差校正来提高股票收益率预测精度的新思路。首先,利用训练样本构建灰色神经网络模型,然后对股票收益率进行初步预测;其次,引入EGRACH模型来挖掘和分析预测误差序列的内部信息,并对该序列后续点进行预测;最后,利用误差预测结果对股票收益率的初始预测值进行校正。文章以上证综合指数数据为例进行分析,结果显示,与校正前的预测精度相比,校正后的预测精度提高了9.3%,表明EGRACH的误差校正过程是有效的,也验证了该方法的可行性。

Stock Returns Prediction Based on Error-Correction Grey Neural Network

In the stock market, it is a crucial issue concerned by all market participants to predict stock returns accurately. However, due to the complicated factors affecting stock market, experience shows that it is very difficult to improve the accuracy of forecasting by setting up single forecasting model. In this article, the deficiencies of the present methods for stock returns forecasting are described and a new approach on forecasting stock returns with the improvement of prediction accuracy is proposed by performing error analysis and correction. First, the grey neural network forecasting model is established by using the training sample data, and then it is used to carry out the preliminary prediction of stock returns. Second, the EGARCH is introduced to analyze the prediction error sequence and predict the subsequent error. Third last, the correction of preliminary prediction values is calibrated. The simulation calculation of the Shanghai composite index as an example is performed and the results show that the precision of prediction is increasing by 9.3 percent significantly compared with the prediction accuracy before correction. Researches also suggest that the error correction process is valid, and then thus the feasibility of this method is verified.