摘要
用微熵率法求得相空间重构的最优嵌入维数及时滞,应用最优嵌入维数及时滞对一维汇率数据进行延时嵌入相空间重构。然后,应用卡尔曼滤波算法在重构后的相空间中对汇率系统进行建模与预测。实验结果与遗传(GA)神经网络预测进行了比较,实践表明,该算法在短期汇率预测中,速度及准确率上均优于GA神经网络。
The optimal embedding dimension and the delay time are obtained by differential entropy ratio method. The phase space of the one-dimension exchange rate is reconstructed by the optimal embedding dimension and the delay time. The prediction model of the exchange rate system is established with the Kalman in the reconstructed phase space. Compared with the genetic algorithm (GA)neural network predicting model,the proposed method is better in prediction speed and accuracy of shortterm exchange rate.
出处
《计算机应用与软件》
CSCD
北大核心
2008年第8期79-80,107,共3页
Computer Applications and Software
基金
福建省自然科学基金(A0540005)
关键词
短期汇率预测
相空间重构
微熵率
卡尔曼
Prediction of shortterm exchange rate Phase space reconstruction Differential entropy ratio method Kalman