FERARI Architecture in Call Anomaly Detection

Big Data platform developed by T-Mobile Poland (TMPL) enabled HT to pilot implement smart network monitoring in order to detect anomalies in call performance and indirectly affect customer experience.

The pilot project revolved around “silent” calls. HT is struggling with customer complaints for the last several months with no successful problem resolution. There are only few systems capable of decoding speech and allowing analysis of “silent” calls and their implementation requires very big investment. TMPL Big Data platform uses heuristic approach based on analysis of user behaviour.

Input data requirements for this use case were also anonymized CDRs from the whole network. A total of 70 million call records were analysed. Analysis was executed on Teradata Aster discovery platform.

Analysis showed that there is significant number of “silent” call events in HT network. There is a trend based increase of “silent” calls in given cells which can predict network quality disturbances and can predict some of them. Day based peaks can indicate network problems.

Silent call analysis
Results of the “Silent” call analysis in HT

The project discovered a lot of cells with significant daily average number of “silent” calls which require additional analysis. Location based reports showed differences in “silent” calls per region in Croatia.

Trial conclusions:

  • TMPL trial results show valuable information about silent calls affected cells (high level view);
  • Detailed analysis (daily trends and peaks, per day and per cell) could be provided further, assuming project is set up;
  • Next steps in optimization can be undertaken already; and
  • Based on trial results, more detailed analysis and other use cases would be beneficial to HT.

Additional processing of the data concluded that affected cells spread all over the territory and that femto cells, tunnels, cities, coast are more affected (drop calls combined). Due to this it could be beneficial to implement FERARI architecture in order to handle this data spread, i.e. information from all the cells in the country.

These potential use cases should be able to handle a vast amount of data (CDRs, network information, etc.) and therefore it should be beneficial to implement FERARI architecture to handle these types of events in a fast and efficient manner.