Abstract: Based on analyzing the shortcoming of present traffic flow forecasting, the maximum likelihood estimationwasused to forecastthe traffic flow ofroad sections. The essence of thismethod is thatupstream observationalvariables in a successional time region are looked as explanatory variables, and the maximum likelihood estimation is applied to seek the relationship between the observational variables and the variables to be predicted to forecast the traffic flow of a road section. An example shows that themaximum error rate between the forecasted and actual values of traffic flow is 5. 76%.