Optimized BP neural network algorithm for predicting ship trajectory

• Published in: 2020 IEEE 4th Information Technology, Networking, Electronic and Automation Control Conference (ITNEC) Vol. 1; pp. 525 - 532
• Main Authors: Ma, Shexiang, Liu, Shanshan, Meng, Xin
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.06.2020
• Subjects: AIS; Biological system modeling; BP neural network; Genetic algorithm; Genetics; Navigation; Neural networks; Particle swarm optimizationt; Prediction algorithms; Predictive models; Ship track prediction; Trajectory
• DOI: 10.1109/ITNEC48623.2020.9085154

Ship navigation trajectory prediction is very important for ship transportation service. Therefore, a BP neural network based on ship's AIS data is proposed to predict ship trajectory. Aiming at the random characteristics of BP neural network initial weight threshold and the characteristics of easy to fall into local minimum, GA (Genetic Algorithm), PSO (Partical Swarm Optimization), ACO (Ant Colony Optimization), DE (Differential Evolution) and GA-PSO are used respectively to optimize the BP neural network. The experimental results show that the five optimized BP neural networks can not only fully extract the nonlinear features of the data, but also have higher prediction accuracy than the traditional prediction methods, and the prediction accuracy of the GA-PSO-BP model. The highest, the mean square error (MSE) of the overall navigation trajectory, navigation longitude and navigation latitude are 7.6638^{\ast}10-6, 4.7618^{\ast}10-6 and 1.0566^{\ast}10-5 respectively.