Optimization of GM(1,1) Prediction Model Based on Background Value and Its Application

According to the drawback of the background value in GM(1,1) model, a new method for computing background value is proposed which uses the non-homogeneous exponential law to fit the accumulated sequence in order to get the area where the practical curve is on as background value. Experiment by using traffic data of Shenzhen between 8:00 to 8:50 on October 9 and 10 in 2007 shows that fitting precision and prediction precision of the new reconstructing background value of GM(1,1) model are improved.