GM(1,1) Model Based on Least Absolute Estimation of Residual
Based on the choice characteristic of least absolute estimation ,and the principle of the minimum of summation of residual absolute value,the GM(1,1) model is proposed to distortion inspecting and forecasting.Exemples show the precision of least absolute estimation is uncertain to be higher a little than that of least squares estimation when observation value contains gross error,and the precision of least absolute estimation is lower a little than that of least squares estimation when observation value doesn’t contain gross error.So least absolute estimation isn’t beter to GM(1,1).
Y. H. Deng "GM(1,1) Model Based on Least Absolute Estimation of Residual", Applied Mechanics and Materials, Vols. 90-93, pp. 2881-2886, 2011