At last week’s Cable Congress, Tele2’s chief architect, Rasmus Aveskogh, outlined the firm’s strategy to leverage the data mining technology it acquired along with Swedish service provider Com Hem.
Tele2 is building a streaming analytics and big data platform using the Com Hem developments.
“We brought these capabilities inhouse to identify anomalies, which we associate with a QoE (quality of experience) timeline and then assign an anomaly a magnitude via QoS (quality of service) models. This then produces reports on production goals, losses and yields – quantified in customer experience whether WiFi issues, capacity issues or problems in the network itself,” said Aveskogh.
This new approach has been introduced to support Tele2’s aggressive transition to the CCAP (Converged Cable Access Platform) technology, which caused a spike in customer service calls. Tele2’s response was to gather data directly from all modems and then check spectrum manually, but this was an arduous and time-consuming process.
So, this gave rise to the second phase in the operator’s network evolution – bringing in machine learning techniques to find anomalies in the data and then apply labels to these samples (such as LTE ingress, GSM ingress, tilt, ripple). The labelled data is enhanced with algorithms to quantify anomalies and predict QoS. Aveskogh gave the example of projecting new single carrier QAMs, allowing engineers to see customers who were viable to receive SC-QAM signals and from there evaluate how to proceed.
In February 2019, Tele2 reported its lowest call ratio in six years, and the figure has apparently fallen further since.