Clustering of Anonymized Users in an Enterprise Intranet Social Network

Abstract

Modern enterprises aim to implement increasingly efficient working methods. One of the trends is the use of process knowledge to guide users about the proper way of performing specific tasks, getting rid of the misleading influence of inexperienced users or veteran users with incorrect habits. But implementation of this kind of working methods can give rise to conflicts and general stressful interdepartmental atmosphere within the company. To this end we are designing a system that allows user ordering based on the amount of daily work done so that more proficient users may have a greater influence in their group, helping all the members of the group to a reach higher levels of performance. A key feature of the proposed system is that the users are anonymized at all times. This measures seeks to avoid personal feelings to interfere on the acceptance and generation of working recommendations. Even the number of user groups and their components are unknown at all times, so that the recommendation system is intended to be fully autonomous and self-managing, increasing overall work efficiency by using the experience of some workers but without disclosing the identity of those employees. The aim is to obtain the personal alignment of the user with the proposed recommendations. We report evaluation of the approach on two test cases in a real business environment, where it has been observed that the proposed system is capable of correctly clustering users and identify the most proficient within each group.