Skip to content

Devoteam develops AI tool that provides real-time insights for Belgian telecom operator

What if big telecom operators could rely on artificial intelligence (AI) to make predictions about the future stability of their digital TV service? It became reality for a Belgian telecom company with the help of Devoteam. Today, the telecom operator is automatically notified when a degradation of the service is detected. What’s more, thanks to the AI model, the company is gaining insights about correlations between different issues.

The full digital TV service consists of many different components. From the individual TV decoders at the customers premises, to different back-end systems and firewalls. All these components generate a wealth of data that is collected centrally. This data paints a picture about the operational state of the full service.

The telecom operator is no stranger to the value of data analytics. In fact, they’ve been using some forms of business intelligence or more specialized analytics tools for years. However, the large amount of customer data that is collected cannot be analyzed quickly. Moreover, analyzing this data is a complex operation that often is a manual endeavor. Therefore, a new approach is needed, that automatically analyses and interprets the incoming data. This would greatly reduce the workload of the TV operations team. That’s why the telecom operator reached out to Devoteam to develop a model that is able to compare the incoming data stream, in real-time, with historical based predictions.

The challenge

It was of utmost importance that the model could quickly identify issues and provide insights in real-time. The full data pipeline, from data generation to data insights, should not take more than a couple of minutes. Catching and solving issues as soon as possible is the key objective of the TV operations team. This approach will help increase the customer satisfaction and reduce the impact on for instance the call centers. Furthermore, a clear overview of the current state should always be available in one concise dashboard. If the system identifies critical situations, the engineers should be alerted automatically. What’s more, false alarms should be avoided since the experts’ time is costly.

An important, and challenging, characteristic of the data originates from the usage behavior of the clients. The load on the TV platform consists of multiple varying patterns that need to be taken into consideration. From short term changes between morning and evening, to seasonal changes. A small number of errors in the morning might be cause for alarm, while the same level in the evening might be expected. 

The solution

To implement the required solution, our experts started collecting, pre-processing and validating the required data. Data quality issues were reported and able to be solved early in the process. This serves as solid foundation for the current, and potential future solutions.  

Once this foundation was in place, the Devoteam experts started building the AI models that capture the different patterns present in the data. Based on these models, the experts were able to create predictions about the future stability and behavior of the service. Devoteam then integrated the AI tool into the existing architecture and set up the necessary data flows for the input and output of the system, as well as the continuous retraining of the model.

Main benefits

This improved approach for monitoring the TV service, led to the following benefits.

  • Automatic notifications on anomalous behavior, saving lots of manual hours.
  • Gaining insights in the correlations between issues, which helps the root-cause analysis of some problems.
  • Gaining insights on how a new software version affects the general state of the STB-network.
  • Detection of smaller anomalies. Previously, some ‘small’ problems were not detected manually. The early discovery with the new approach, gives the telecom operator more time to work out a solution for them. This prevents theses errors to grow into a service disrupting problem that affects many end-users.