Customer Experience, Telco, Machine Learning, Automation… all sounds a bit dry, no? If not, this should help shed light on a specific use case, but also provide details and lessons learned while building the solution. If so, read on to perhaps find some surprising details. Many are aware of the impact from Machine Learning but may not be aware just how many portions of the enterprise it is now altering -- or the magnitude of those changes.
With building an ML platform in mind, specifically to improve and automate the customer experience, this paper will illustrate how we accomplished this. It will also detail lessons learned and insights not only on the data itself, but on the architecture and thinking behind it.
Another focus is the challenges and differences of scaling and maintaining the Machine Learning components of the architecture. The highlight here is that while Proactive Network Management (PNM) has laid out a number of excellent methods on how to collect and analyze network telemetry from CPE and other headend equipment, it has not covered the aspects of what to do or how to handle the data with respect to utilizing ML models.
This paper will begin with the Customer Experience use case, and work from that point through to the end solution. A number of open source technologies are referenced as possible implementations for components. Other components can certainly be used.
The desire to share this information stems from the fact that some of these solutions are not easy. They require not only multiple resources to develop the front-end user interface, but the back-end platform as well. After the ML models have been trained, the task of building the platform, scaling, testing specific tools and gluing everything together is still rather manual and prone to performance and optimization challenges. With those things in mind, the details throughout this paper should aid the reader in determining directions to go when considering, building and scaling such a system.