Meta-NWDAF: A Meta-Learning based Network Data Analytic Function for Internet Traffic Prediction

With the coming of the B5G and 6G era, lots of research reports held on predictions that the number of connected devices will keep exploding. According to specifications determined by the 3rd Generation Partnership Project (3GPP), when excess devices request internet may lead to signaling overhead of the charging system, it is necessary to predict the internet traffic required. Therefore, we take the advantage of meta-learning to effectively predict according to a few samples in the past. We implement the network data analytics function (NWDAF) and charging function based on meta-learning and implement it on a public cloud platform. Experimental results show that our proposed Meta-NWDAF architecture can reduce signaling significantly. Our research contributions are to show that meta-learning can be applied to not only classification problems but also time series prediction problems, and we also prove that the future diverse connected devices are well suited for leveraging me-ta-learning. The managerial implication of this research is that our proposed architecture effectively reduces the signaling overhead for the charging system.