To improve the accuracy of urban traffic flow state identification under limited data condition

a model combining grey clustering and fisher transform (GC-fisher model) is proposed. First

the parameters of traffic volume (Q)

speed (v)

and occupation (O) were divided into four categories according to the grey clustering theory. Then training dataset was established by using the classified data

and the fisher transform and discriminant functions of each traffic flow state were obtained by using the GC-Fisher model. Afterwards

the combination of K-means and multi-class support vector machine (K-SVM) model was selected as the comparison model. Results of model discriminant rate and the case study show that the GC-fisher model outperformed K-SVM model under limited data condition: the discriminant rate was 92% by using GC-fisher

while the rate was 69% by using K-SVM. The GC-Fisher model can improve the discrimination accuracy of traffic flow state under limited data condition.