Container volume forecast of inland river port considering collection, distribution and cluster analysis

The forecast of container volume is of great significance to the planning of inland river port collection and distribution.Basing on the "Shuttle Bus" barge collection and distribution system of Guangzhou Port, combined with the regional characteristics of container collection and distribution volume of terminals in the Pearl River system, selecting container terminals in Beijiang Tributaries to conduct an empirical study.Firstly, cluster analysis was carried out on the target terminal group to determine the number of variables under the optimal prediction of random forest characteristic variables of the collection volume, distribution volume and the total traffic volume.Comparing the Mean Absolute Error（MAE） of the prediction results, which are directly predicted by throughput volume, collection and distribution volume, and combined with cluster analysis.The effects of cluster analysis and separate prediction of lumped and dispersed volume on the results of random forest regression of inland port throughput are discussed.Finally come to the conclusion that the results of cluster analysis are more accurate when considering aggregation and distribution.This empirical study can also provide methodological support for the container volume prediction of other tributaries of the Pearl River system.