Bernard Burg spearheads Comcast’s applied AI team at Comcast Silicon Valley. Our mission is to improve the user experience by monitoring the behavior of our networks, predict their performance and implement self-healing. Bernard earned a PhD in machine learning with a focus on behavioral process control, a predecessor of reinforcement learning.
Predicting and Preventing Avoidable Truck Rolls (ATR)
When call centers cannot solve issues, they roll a truck to customer dwellings. Experience shows that a significant number of these truck rolls could be avoided as the root cause of the issue might be in the network upstream of the dwelling. This presentation describes the method we created to build a solution capable of predicting these ATR with a very high accuracy. The machine learning algorithm uses both network health indicators as well as the truck operation records. This method is the first of its kind and is currently being tested for deployment.