基于改进QPSO算法优化SVR的上证指数预测

Prediction of Shanghai Composite Index Based on SVR Optimized by Improved QPSO Algorithm

研究证券指数预测问题,证券指数变化具有时变性、不确定性、非线性等特点,传统的线性预测方法无法正确反映证券指数的变化规律,且神经网络存在局部极小值、过拟合、泛化能力不强等缺陷,预测精度比较低。为了提高证券指数的预测精度,提出了一种利用改进量子粒子群(QPSO)算法优化SVR的证券指数预测方法(改进的QPSO-SVR)。首先采用改进的QPSO算法优化SVR的参数,然后将优化后的SVR对证券指数的非线性变化规律进行建模。最后选取2002年10月18日到2004年1月9日上证指数的历史相关数据进行实验。实验结果表明,采用改进的QPSO-SVR提高了证券指数的预测精度,减少了预测误差。预测结果能够为股票投资者提供有价值的参考意见。

Study Shanghai composite index forecasting based on Support Vector Regression （SVR）. This paper proposed a stock index forecasting method based on SVR optimized by improved quantum - behaved particle swarm optimization（QPSO） algorithm（ Improved QPSO- SVR）. Firstly, the parameters of SVR were optimized by the im- proved QPSO algorithm. Secondly, the optimized SVR was used to establish a prediction model. The model reflectes the change rule of the stock index. Finally, the historical data of Shanghai composite index from October 18, 2002 to January 9, 2004 were used for simulation test. The experimental results show that, the improved QPSO - SVR can improve the accuracy of the stock index forecasting and reduce the prediction error. The forecasting results can provide valuable advice for stock investors.