基于误差校正的ARMA-GARCH股票价格预测

针对时间序列预测和简单回归预测各自的侧重点不同,综合两者优点,对股票价格进行预测。首先将股价数据转换成对数收益率,利用ARMA-GARCH模型对收益率序列建立模型,对上证指数股票价格进行初步预测;然后建立回归模型对GARCH模型误差中未被解释的成分进行分析和拟合,利用回归模型预测的误差对GARCH模型预测结果进行校正。在选择回归模型变量时,引入变量间的相关性分析筛选合适的影响因子,利用主成分分析方法提取影响因子中包含的信息,实现对解释变量的降维,获取具有代表性的综合指标,以提高建模精度。实例研究证明该方法对于上证指数股票价格预测较为准确。

Stock Price Forecast with ARMA-GARCH Based on Error Correction

This paper,based on each different focus of the time series prediction and the simple regression prediction,integrates the advantages of them and predicts the price of the stock.First of all,the price data is converted to logarithmic rate of return and the ARMA-GARCH model is used to build a model on yield sequence which can make the preliminary forecast on the Shanghai Composite Index of stock prices.Then the Regression model is built to analyze and fit the unexplained component in the error of the GARCH model and correct the result predicted by the GARCH model.As selecting the regression model variables,correlation analysis between variables is brought in to pick out appropriate impact factors whose information is extracted out by the principal component analysis to achieve the dimension reduction on explanatory variable and then obtain the typical composite indicator in order to improve modeling accuracy.Examples of studies have shown that the method for predicting stock prices on the Shanghai Composite Index is more accurate.