Assessing the Impact of Network Events with User Feedback

Abstract

User feedback data, generated when users call customer care agents with problems, is a valuable source of data for understanding network problems from users' perspectives. However, this data is extremely noisy. In this paper, we design a framework, LOTUS, to assess the user impact of network events from the user feedback, through a novel algorithmic composition of co-training and spatial scan statistics. Through experimental analysis on synthetic and real data, we show the accuracy and practical nature of LOTUS.