Machine learning techniques applied in activity and action prediction for an emphatic space using context

• Main Authors: Bautista, Nikka Jennifer G., Cua, Manuel M., Jr., Gonzales, Jed Aureus J., Urquiola, Marc Angelo B., Trogo-Oblena, Rhia S., Suarez, Merlin Teodosia C.
• Format: text
• Published: Animo Repository 2009
• Subjects: Human activity recognition; Pattern recognition systems; Ambient intelligence; Human-computer interaction; Context-aware computing; Computer Sciences; Physical Sciences and Mathematics; Software Engineering
• Online Access: https://animorepository.dlsu.edu.ph/faculty_research/12844
• Institution: De La Salle University

Activity prediction is an integral part in the field of empathic computing though, in recent years, it has been subject to intense scrutiny due to the immense complexity of the task. An activity is composed of a set of actions however, due to the nonlinear nature of actions, it is difficult to identify the marker as to when the set of action for an activity begins and ends. The segmentation of actions is an integral part of activity recognition, and subsequently activity prediction, due in large part to an activity being defined as a sequence of specific actions. Several studies have seen success in accurate activity recognition, although only few have accomplished accurate activity prediction. This paper presents different supervised and unsupervised learning techniques and their respective results using data that has been gathered in the Empathic Space, TALA.