基于ARIMA-GARCH模型的股票价格预测

Stock Price Prediction Based on ARIMA-GARCH Model

摘要: 本文引入了误差校正的思想，先利用ARIMA-GARCH模型对日收盘价进行初步预测，但是预测精度不高，通过对误差序列进行的白噪声检验，发现误差序列存在还有未被ARIMA-GARCH模型提取的信息。再利用变量间的相关关系，寻找与误差序列相关的变量。随后通过主成分分析对变量进行降维，以筛选出合适的解释变量。将解释变量与误差序列进行回归建模，对误差进行预测。最后将预测的误差值与ARIMA-GARCH模型的预测值相加，得到最终的预测值。通过将最终的预测值与ARIMA-GARCH模型的预测值相比较，观察到预测精度有了很大的提高，进而验证了引入误差校正的方法是合理的。

Abstract: This paper introduces the idea of error correction. Firstly, ARIMA-GARCH model is used to predict the daily closing price, but the prediction accuracy is not high. Through the white noise test of the error sequence, it is found that there is information in the error sequence that has not been extracted by ARIMA-GARCH model. Then the correlation between variables is used to find the variables related to the error sequence. Then, the dimensionality of the variables is reduced by principal component analysis to screen out the appropriate explanatory variables. Regression modeling is carried out between explanatory variables and error series to predict the error. Finally, the predicted error value is added to the predicted value of ARIMA-GARCH model to obtain the final predicted value. By comparing the final prediction value with the prediction value of ARIMA-GARCH model, it is observed that the prediction accuracy has been greatly improved, and then it is verified that the method of introducing error correction is reasonable.