Unsupervised work knowledge mining through mobility and physical activity sensing

Abstract

Knowledge of working professionals gained through years of experience is invaluable for any organization. Extracting this knowledge allows an organization to optimize internal processes and facilitate training of new hires. Therefore, there has been a significant research effort in developing techniques for automated knowledge mining at workplaces. However, research in the past have been focused mainly on extracting knowledge of stationary professions such as office workers, who perform most of their day-to-day tasks at their desk. In this work, we propose an approach for mining work knowledge of physically active professions such as nurses, firefighters, waiters, housekeepers or janitors. We leverage the advances of mobile sensing to extract knowledge from workers with high level of mobility and physical activity patterns. We demonstrate the feasibility of the proposed approach on a real-world scenario of a janitor as a study subject. We show that using data collected from mobile devices carried by a janitor throughout their work, we are able to extract knowledge rules that describe generalized patterns of janitor's behavior. We expect the proposed method to be applied to other fields to mine knowledge from workplaces.