AI detects problems in the Beeline network
Since the beginning of the first quarter of 2023, Beeline has completely switched to an operational scheme for detecting errors on the network using the algorithms of the Virtual Quality Expert system, which during this period detected about 3.5 thousand problems that potentially affect subscriber service.
About 1,500 such cases are known and have been studied previously, while the remaining 2,000 are errors due to problems on the transport network that could affect customer service and increase the number of calls to the help desk.
Network Operations Director Aleksey Kazaev : “About 70% of all identified problems were ahead of subscribers’ complaints by more than 2 days, another 30% were indicated on the same day, or ahead of the day. About 60% of the problems that were predicted by the algorithms of our system were resolved before the receipt of subscriber complaints. The rest turned into local errors and was accompanied by client requests, which were processed in priority order.
“Virtual Quality Expert” automatically monitors the operation of networks, fixes possible deviations, informs the relevant specialists about them and controls the elimination process. In addition, based on machine learning algorithms, the system can predict the number of requests from clients for a particular incident in order to minimize its impact on the quality of the network.
Based on an assessment of the volume of identified problems and their classification, the so-called “health level” is calculated both for the entire beeline network and for each branch separately. The system “monitors” each of hundreds of thousands of network elements, automatically detects problematic base stations , classifies them according to eight types of impact on subscriber service and groups objects with a single problem. Machine learning algorithms predict the potential number of customer referrals based on the prior year’s history of matching complaints and bugs. ■