VIRTUAL NETWORK FUNCTION RESOURCE DEMAND PREDICTION METHOD ADOPTING MACHINE LEARNING

摘要

The present invention relates to a technology performing prediction using a machine learning method in performing a resource demand prediction of a VNF which provides a core technology in a network virtualizing environment. The present invention, in order to predict VNF resource information, does not use only a resource of an individual VNF but uses only information of directly or indirectly related surrounding VNFs such that machine learning consumption time can be reduced corresponding to an SFC size and an overall network size. In addition, the present invention uses the machine learning in order to predict VNF resource information and uses a target dependent LSTM model among a plurality of machine learning models, wherein the model can perform machine learning for one data object itself. In addition, the learning model according to the present invention performs attention based machine learning to selectively determine core information in performing the machine learning, and operates attention learning in an aspect side using aspect embedding such that the machine learning can be set to be centered on a target VNF.;COPYRIGHT KIPO 2020