Vessel Traffic Flow Prediction Using LSTM Encoder-Decoder

Abstract

Accurate vessel traffic flow prediction is of vital indispensable for the enhancement of capability of navigation, the optimal allocation of port resources and the improvement of the navigation safety. In order to improve the accuracy of vessel traffic flow prediction, a prediction approach based on Long Short-Term Memory Encoder-Decoder (LSTM-ED) is proposed for multi-step prediction of vessel traffic flow. In order to study vessel traffic flow, a statistical approach of vessel traffic flow is proposed by using the data of Automatic Identification System (AIS). The vessel traffic flow data of the Liuhe Waterway in the Jiangsu section of the Yangtze River is selected as the experimental data for model training, validating and testing. Build the LSTM-ED-based prediction model and verify its validity in predicting vessel traffic flow. Experimental results show that the proposed prediction approach can accurately predict the trend of vessel traffic flow, the LSTM-ED-based approach can obtain better prediction performance compared with other baseline methods.