Prediction of ship traffic flow based on RF-bidirectional LSTM neural network

Abstract

To increase the proportion of vessels entering and leaving the port, we will improve the accuracy of vessel traffic flow forecasts to meet the future development needs of the port. This paper proposes a method for predicting ship traffic flow based on RF (Random Forest, RF)-bidirectional LSTM (Long Short-Term Memory, LSTM). In this paper, the random forest (RF) algorithm is combined with LSTM and two-way LSTM to make predictions and comparative studies, and apply it to the 48-month forecast of the total number of ships entering and leaving the port in Qingdao Port from 2016 to 2019. The results show that the method based on RF-Bidirectional LSTM has the highest prediction accuracy, and compared with the other two prediction models, its evaluation index root mean square error, average absolute error and average absolute percentage error are 167.49, 95.27 and 3.64%, respectively. Based on RF-LSTM neural network has the lowest prediction accuracy. The prediction accuracy based on RF-LSTM neural network is the lowest. The forecasting method of ship traffic flow proposed in this paper is expected to provide decision-making guidance for the future development and planning layout of the port.