應(yīng)用預測最小二乘準則的AR模型判階方法
STUDY OF AR MODEL ORDER SELECTION APPLYING CRITERION OF PREDICTIVE LEAST SQUARES
-
摘要: 本文通過多種AR模型的判階準則的比較提出了應(yīng)用最小二乘預測誤差的滑窗預測最小二乘(Sliding Window Predictive Least Squares, SWPLS)判階準則.采用這種準則的主要優(yōu)點除了準確的判階性能外,對于時變AR模型具有良好的跟蹤特性,同時算法容易實現(xiàn)在線實時處理.文中主要對時變模型參數(shù)和時變模型階數(shù)的多種情況進行了判階模擬,驗證了文中提出的滑窗最小二乘預測判階準則的有效性.
-
關(guān)鍵詞:
- 模型判階; 滑窗數(shù)據(jù); 格形結(jié)構(gòu)
Abstract: A criterion called Sliding Window Predictive Least-Squares (SWPLS) order selection criterion is presented in this paper by comparison of several typical criteria of AR model order selection. In addition to property of fine order selection the criterion has advantage of good tracking property and can be easily implemented on-line in real time. The effectiveness of SWPLS criterion is verified by simulation experiment of order selection for models of time-varying parameter and time-varying order. -
計量
- 文章訪問數(shù): 2220
- HTML全文瀏覽量: 176
- PDF下載量: 385
- 被引次數(shù): 0