Beeline specialists have developed the “Virtual Quality Expert” system.
It detects instances of observed network performance, captures possible deviations, reports them to specialists, and looks for an engagement process.
This system analyzes more than 30 network parameters around the clock and, based on machine learning algorithms, can predict how many requests are received from clients for a given capture or another. Based on a history of complaints and network degradation over the past year.
Based on the assessment of carefully identified problems and their exclusion to determine the “level of health” for the entire beeline network and each branch in the exclusions. The system “monitors” each of the hundreds of thousands of basic elements, automatically identifies problematic base stations, classifies them according to an eight-type survey for subscriber service, and groups objects with a single problem.
“Virtual Quality Expert” is our development task, which identifies problems even before the client notices them and leaves us a request for a deterioration in the quality of service. In practice, the training model remembers which values of the statistical indicators and to what extent the characteristics of the customer experience.
This makes it possible to predict the scale of problems and failures in the network with high caution, as well as to track resources to solve them, which is extremely important today. The system helps to maintain and improve the quality of the network, and, as a result, reduce the number of locations where low-quality services are connected.
— Alexey Kazaev, Director of Network Operations, PJSC VimpelCom
For a year of operation, the system receives more than 2.5 thousand complaints with a delay of 70%. They are expected to work with data that happens hourly. This improves the accuracy and speed of predictions, and therefore reduces the time to identify and fix problems.
Source: Iphones RU
