Grey Verhulst Model based on TLS and GA*

摘要

Aiming at error of fitted value causing by the identification parameter and initial value of Verhulst model, first, Total Least Squares (TLS) is used to improve the identification parameters which cause the error of fitted value. Secondly, improved Verhulst model is built up via identification parameters by TLS and boundary value correction by GA under minimizing the sum of the squared error. Numerical example results in ex post testing stage: relative percentage errors (RE) of Verhulst model are 10.76%, 9.43%, and 6.31%, respectively. And, relative percentage errors (RE) of improved Verhulst model are 1.41%, -0.27%,and -3.99%, respectively. The results of example show that the fitting accuracy and forecasting accuracy of improved Verhulst model are better than the traditional Verhulst model.