Abstract:

To improve the classification performance of the algorithm on AIS trajectory data, and combine multi-participants data with security data mining in the process of federal training, this article proposed an algorithm named ship AIS trajectory classification algorithm based on federated random forest. By integrating the BCP homomorphic encryption algorithm to design a protection with average privacy-preserving, it solved the problem of multi-participants securely training decision tree on federated learning. The algorithm analyzed the ship’s trajectory data and extracted the optimal trajectory features, which could be used as the input of the model. It realized the federal classification of four typical ships, namely fishing boat, passenger ship, cargo ship and oil tanker. Further experiment from two aspects of accuracy and efficiency shows that, besides its security advantage, the algorithm performs well in terms of classification effect, reduces the computation overhead of the participant client, and realizes the security and federal data mining by multi-participants. At the same time, it can be applied to ship trajectory identification and ship navigation risk analysis.

Key words: federated learning, AIS trajectory, random forest, homomorphic encryption