Precisely forecasting the container demand usually plays a decisive role on government investment in constructing port container wharves and inland transportation systems. It is particularly important for an import/export oriented country. Regression models are commonly used in forecasting the container demand in Taiwan Area, however, such models often violate the basic assumptions of regression analysis which may lead to biased and/or inefficient estimations. Forecast errors are therefore often greater than 40%. This paper constructs a regression model by investigating the vital assumptions. Details of the model estimation procedures are presented. In order to prove the forecasting power, the statistical data of 1985-1989 container volumes in Taiwan Area are used to compare with the theoretical values forecasted by this model and by other previous models. It shows that the maximum forecast error of this model is less than 9%, which is superior to the previous models.