Prediction of Telecommunication Network State Based on Knowledge Graphs

Abstract

The article focuses on developing a new prediction method for telecommunications network state based on knowledge graphs. The article analyses the prediction methods supported by current network monitoring and management systems. The proposed prediction method is based on analysis of behavior of each end-user device that allows reach high accuracy of prediction. Knowledge graphs of the networks are build using domain knowledge and statistical data. They contain knowledge about the possible states of the elements of the networks along with the knowledge about possible transitions between them that can be presented in the form of prediction rules. These rules allow predict the state of the network for the defined timestamp in the future. In the case study the practical task of internet traffic prediction for a segment of cable TV operator network is considered. Conclusions are formulated, and the areas of further research are defined.