Sponsored by: Juniper Networks

Abstract

Juniper Networks is an American company selling networking hardware and software. When a customer needs assistance with a product, a customer support case is opened. The case can go through many stages before being resolved. The route a case takes to get to the resolver is currently human-controlled. This project aimed to assist case routing through the use of machine learning. We developed a system that can use a pipeline of data including previously resolved cases to model the relationship between a case and the engineer responsible for solving the problem. We used a deep learning neural network to create a system that could predict both the resolving group and engineer achieving up to 96% accuracy when predicting the 10 most likely groups, and 42% when predicting the 10 most likely engineers. We built a web application for future demonstration of this system.

Download

• Final report: pdf
• Slides: pptx