Abstract: We generalize wholly social containerized freight volume from 1995 to 2012. We also screen out indexes with gray correlation analysis method.We further construct BP neural network prediction model, based on which we predict wholly social containerized freight domestic demand from 2013 to 2020.Prediction results show that BP neural network model has better fitting effect for the actual past value and little error. It demonstrates that the model has higher reliability and practicability.

Key words: containerization, freight volume forecast, BP neural network