Uncertainty identification in context-aware systems using public datasets

Abstract

Uncertain situations are naturally embedded in intelligent environments due to a series of factors. Changes in users’ behaviour and the lack of context data are some of the reasons for this phenomenon to happen. Computing solutions for such domains should consider strategies to cope with the uncertainty problem, once it may influence the behaviour of services and, consequently, affect the way users interact with the environment and with the system. This paper presents the validation of an approach to tackle this obstacle. The main goal is to provide subsidies for the identification of uncertainty. The proposal includes the definition of a decision tree to classify context data and use it to reduce the level of uncertainty while building situations. The conduction of the experiments was through case studies created based on two public datasets containing Activity Daily Living data from smart houses.