Abstract:To solve the morbidity problem of an extended grey prediction model in the process of parame-ter identification,the morbidity degree of the model was calculated by the condition number of matrixspectrum,and the main factors that affect the morbidity of the model were summarized. It was concludedthat there was no serious morbid character in some cases. Finally,the results of an example showed thatif the numerical value of original data were too large,the significant disturbances in the accuracy of themodel that caused by the small error of the data could be avoided by multiplying transformation on originalsequence.