Partner: T-Mobile Czech Republic a.s.
Deutsche Telekom-IT
Competence & Delivery
Centre Prague – CDCP

Field: telecommunications

In cooperation with the competence center for Deutsche Telekom networks development operated by T-Mobile
Czech Republic a.s., we are involved in performing analysis of data generated from the backbone network of the mobile operator in order to increase reliability and reduce cost in maintaining new technologies ensuring 4G and 5G mobile networks operation.

High reliability of telecommunication networks is ensured by sub-analyses performed for each technology or network unit. However, a number of technical problems in networks are caused by a combination of their various parts or technologies. Therefore, it is difficult to identify the causes of these problems using the existing approaches and methods. Moreover,
identification of the root cause is also time-consuming. The objective of the collaboration is to find the key data
sources, gather information about technical problems in one place and identify performance indicators, which
can be used to increase reliability and prevent problems in the network. The results will be verified in a testing environment, which must be compatible with the production environment available to the industrial partner.

The joint research aims to analyze the data obtained from the backbone network of the mobile operator, in
particular from those parts ensuring network rules and billing. One of the main tasks is to work out the procedures followed by analyses both leading
to identification of problems in the implemented voice technology for the Voice over LTE (VoLTE) service,
which stands for Voice over Long Term Evolution. VoLTE supersedes older voice technologies in mobile networks
against which it introduces an enormous conceptual difference. The various alternative implementations of
the technology cause problems with mutual cooperation of the network entities.

Our analysis is focused on basic indicators in the network, such as volume of disconnected calls or reported errors of the used protocol. Based on these
errors, the key performance indicators are proposed. Regarding the nature and volume of the data analyzed, specific tools for big data analyses need to be applied. The current results indicate that most of the problems with the VoLTE technology are associated with particular mobile network terminal equipment and cells, which do not appropriately cooperate. This is caused by unanchored specification of VoLTE and the large number of suppliers of mobile devices operating within a single network. These problematic devices in particular mobile cells have been detected by applying the developed
anomaly detection algorithm, which uses a selected key indicator. An important property of the performed analyses is the opportunity to geographically locate all
events using various data layers, which significantly simplifies their solution on the side of the mobile operator.

Due to the dispersion of polygons representing the model of mobile network coverage, the very display of relevant network layers for analytical purposes presents
a significant technological challenge, despite the progress in public availability of thematic base maps. Based on the past experience, we are preparing data files structures over a longer period to allow time series analyses using the proposed indicators. Pre-processed data sets and processing procedures enable us to use reliable statistical methods such as seasonal and trend breakdown to find anomalies that may appear quite irregularly.


Radim Kalfus
Telco Development team
Competence & Delivery Centre Prague – CDCP
T-Mobile Czech Republic a.s.

For implementation of this solution, we use not only the HPC infrastructure but also the experience of experts in data arrangement conception and design, parallelization of computations, and algorithm optimization. The parameters
obtained from HPC simulation will serve as input for a solution, which shall not require HPC power thereafter. Future integration using machine learning offers us a unique opportunity to automatically detect and classify anomalies in the network as well as to increase the availability of services to end customers by means of timely reconfiguration.