Utilizing Deep Learning for Mobile Telecommunications Network Management

Traditional network and service management methods were based on counters, data records, and derived key indicators, whereas decision were mostly made in a rule-based manner. Advanced techniques have not yet got into the everyday life of operators, mostly due to complexity and scalability issues. Recent progress in computing architectures, however, allows to re-visit some of these techniques - nowadays, especially artificial neural networks -, and apply those in the network management. There are two aims of this paper. First, to briefly describe the possible network and service management tasks within the mobile core, where utilizing deep learning techniques can be beneficial. Secondly, to provide a living example of predicting certain fault-types based on historical events, and to project further, similar examples to be executed on the same architecture. The real-life example presented here aims to predict the occurrence of certain errors during the VoLTE (Voice over LTE) call establishment procedure. The methods and techniques presented in this paper has been validated through live network monitoring data.