A Method to Improve the Performance of Network Data Analytics Function Based on Transfer Learning

How to improve the correctness of network data analytics is one of the study key issues in the study of 3GPP R18. Transfer learning is a method of machine learning aiming at improving the performance and model training efficiency of target AI task, which can be used for reference in network data analytics. The main purpose of this paper is introducing a proposed method to improve the performance of Network Data Analytics function (NWDAF) drawing on the thought of transfer learning. The NWDAF stores the trained ML models in the Analytics Data Repository Function (ADRF) and the ADRF registers the model information to Network Repository Function (NRF). Specifically, to introduce the method in detail, this paper describes the fundamentals, system architecture, key procedures and key parameters of the proposed model transfer method.