In order to improve the forecasting accuracy of container port throughput, an improved dynamic linear model is developed considering all periodic fluctuation factors such as seasonal changes and weather impact. Bayesian estimation and prediction is used to solve the improved model in this research. An empirical analysis of the Nanjing container port throughput is carried out to compare the forecasting effectiveness and accuracy of two models, i.e. the linear growth model and the improved dynamic linear model. Empirical results indicate that comparing with the linear growth model, forecasting results from the improved model are more in line with the actual container throughput. As such, the improved dynamic linear model has better forecasting performance in container throughput of Nanjing Port than that of the general linear growth model. It suggests that the improved dynamic linear model can be used in container throughput forecasting in practice. It also provides references for maritime and port authorities and decision makers to develop proper strategic planning based on more accurate container throughput forecasting.