A Framework for Rank Computation and Aggregation in Fuzzy Environments

Abstract

In many network and IT systems, users submit loosely-defined (or fuzzy) requests to obtain answers, solutions or resources. In this paper, we focus on trouble-shooting problem tickets in an IT service management system. In such a system, problems are typically reported using vague user-generated descriptions of the symptoms (e.g., 'my email is not working!') in a ticket. IT Specialists are then responsible for identifying the troubled component from the reported symptoms. Usually, the ticket is sent to one Specialist at a time until it is resolved hence, an efficient mechanism to select the (next) potential Specialist becomes critical towards a timely resolution of the ticket. In this paper, we introduce a framework to deal with fuzzy requests in a distributed environment. We first show the importance of considering both the traditional graph-based network routing efficiency as well as social network theory to generate important ranking metrics. For each newly arrived ticket, we propose several selection and ranking policies to choose and rank IT Specialists who can potentially resolve the ticket. We evaluate the ranking system with synthetic tickets, generated by an in-depth simulated IT management model that closely mimics an operational system. Our preliminary results suggest the importance of considering both routing efficiency and social connectivity to minimize the resolution time of a ticket.