基于平均包絡線匹配算法的EMD端點效應分析及在股價趨勢分解中的應用
發(fā)布時間：2019-05-07 21:02
【摘要】：經驗模式分解(EMD)能夠有效獲得非平穩(wěn)非線性信號的時頻特征,但傳統(tǒng)的EMD分解算法存在嚴重的端點效應.在深入研究和分析EMD算法的基礎上,提出了一種基于波形匹配的端點效應處理方案,通過計算波形匹配度,在平均包絡線內部尋找與其端部變化趨勢最為接近的子波,并用這段子波代替平均包絡線的邊緣部分,使處理后的平均包絡線極大地接近真實包絡線,并把這種端點效應處理方案的EMD分解算法應用到實際的股票市場價格趨勢分解中.實驗結果表明,與經典的EMD邊界延拓算法相比,本文提出的算法能更有效地抑制EMD分解時的邊界效應,分解得到的固有模式函數(shù)更能體現(xiàn)模擬信號真實的頻率、幅值信息.應用實驗表明:與現(xiàn)有方法相比,該方法更能提高預測精度.
[Abstract]:Empirical mode decomposition (EMD) can effectively obtain the time-frequency characteristics of non-stationary nonlinear signals, but the traditional EMD decomposition algorithm has a serious endpoint effect. On the basis of in-depth study and analysis of EMD algorithm, an endpoint effect processing scheme based on waveform matching is proposed. By calculating the degree of waveform matching, the wavelet which is closest to the change trend of its end is found in the average envelope. This wavelet is used to replace the edge part of the average envelope so that the processed average envelope is very close to the real envelope. The EMD decomposition algorithm of the endpoint effect processing scheme is applied to the actual price trend decomposition of the stock market. The experimental results show that compared with the classical EMD boundary extension algorithm, the proposed algorithm can suppress the boundary effect of EMD decomposition more effectively, and the intrinsic mode function obtained from the decomposition can reflect the real frequency and amplitude information of the analog signal more effectively. The experimental results show that compared with the existing methods, the proposed method can improve the prediction accuracy.
【作者單位】： 華南理工大學經濟與貿易學院;紐約州立大學石溪分校商學院;復旦大學計算機科學技術學院;華南理工大學工商管理學院;
【基金】：教育部人文社會科學研究項目(13YJC790068) 國家自然科學基金(10826053,70871040) 華南理工大學中央高�；究蒲袠I(yè)務費專項資金(2009ZM0081)
【分類號】：F830.91;F224
[Abstract]:Empirical mode decomposition (EMD) can effectively obtain the time-frequency characteristics of non-stationary nonlinear signals, but the traditional EMD decomposition algorithm has a serious endpoint effect. On the basis of in-depth study and analysis of EMD algorithm, an endpoint effect processing scheme based on waveform matching is proposed. By calculating the degree of waveform matching, the wavelet which is closest to the change trend of its end is found in the average envelope. This wavelet is used to replace the edge part of the average envelope so that the processed average envelope is very close to the real envelope. The EMD decomposition algorithm of the endpoint effect processing scheme is applied to the actual price trend decomposition of the stock market. The experimental results show that compared with the classical EMD boundary extension algorithm, the proposed algorithm can suppress the boundary effect of EMD decomposition more effectively, and the intrinsic mode function obtained from the decomposition can reflect the real frequency and amplitude information of the analog signal more effectively. The experimental results show that compared with the existing methods, the proposed method can improve the prediction accuracy.
【作者單位】： 華南理工大學經濟與貿易學院;紐約州立大學石溪分校商學院;復旦大學計算機科學技術學院;華南理工大學工商管理學院;
【基金】：教育部人文社會科學研究項目(13YJC790068) 國家自然科學基金(10826053,70871040) 華南理工大學中央高�；究蒲袠I(yè)務費專項資金(2009ZM0081)
【分類號】：F830.91;F224
【參考文獻】
相關期刊論文 前10條
1 秦賢宏;段學軍;李慧;;基于EMD的我國經濟增長波動多尺度分析[J];地理與地理信息科學;2008年02期
2 張衍廣;林振山;李茂玲;梁仁君;;基于EMD的山東省GDP增長與耕地變化的關系[J];地理研究;2007年06期
3 謝赤;鄭林林;孫柏;張在美;;基于EMD和Elman網(wǎng)絡的人民幣匯率時間序列預測[J];湖南大學學報(自然科學版);2009年06期
4 丁志宏;謝國權;;金融時間序列多分辨率實證研究的EMD方法[J];經濟研究導刊;2009年06期
5 劉慧婷,張e，
本文編號：2471382
本文鏈接：http://www.sikaile.net/guanlilunwen/huobilw/2471382.html
