Comcast is the third-largest Internet provider worldwide, managing massive networks which deliver connectivity and streaming content to millions of customers. Such networks face complex maintenance and troubleshooting issues. We use Machine Learning to analyze and model error patterns to continuously assess the health of our network and ensure a smooth experience for every user. This is supported with a Decision Engine, which can be configured to take appropriate remedial actions such as customer notifications and self-healing directives.
We describe the architecture capable of scaling and handling billions of events per day and explain how H20 helps to implement the underlying learning models. We illustrate the superiority of H2O algorithms in all of the following: accuracy, speed and memory footprint with comparisons to other systems such as Spark ML. #h2ony