Freight volume forecasting is an important part of the reasonable logistics planning. In this paper, a historical freight volume of A company is regarded as research data, mainly basing on the trend extrapolation method and the seasonal decomposition method, establishing a single model and a combination model. Discussion on the time series of short-term forecasting is a direct curve or "remove" seasonal before fitting. For the trend model, there are two choices: One is polynomial model; another is Logistic model. While assuming the influence of monthly seasonal factors is same. We could build four predictive models ultimately, compare the goodness of fitting, then chose the best one. The result demonstrates that the combination model is better than the single model.