Projects per year
Abstract
In this report, we present the final version of the Adaptive Tuning and Personalization (WP5)
component of the M-Eco system. This component is focused on four main areas of interest to
users of surveillance systems: presentation options for recommendation and adaptation, user
and group models, user classification and modeling algorithms, and recommendation, adaptation
and personalization strategies. In each of these areas, we propose improvements over those
presented in the previous deliverable incorporating feedback from medical surveilance experts.
The personalization component is enriched with several extensions. The multi-factor recommendation
model is incremented with more factors, including time decay, tagging activity and
location. Location is also improved with a method to predict event trajectories. Moreover, a
new family of tensor-based recommenders is presented. The tag cloud model is improved with
two new tags selection algorithms. We also introduce a propagation of the (ir)relevant terms
obtained from tag clouds usage to WP3 and WP4 in order to improve their data collections
algorithms. The evaluation part of the report summarizes the benefits and drawbacks of the
models presented and describes possible directions for the future work.
component of the M-Eco system. This component is focused on four main areas of interest to
users of surveillance systems: presentation options for recommendation and adaptation, user
and group models, user classification and modeling algorithms, and recommendation, adaptation
and personalization strategies. In each of these areas, we propose improvements over those
presented in the previous deliverable incorporating feedback from medical surveilance experts.
The personalization component is enriched with several extensions. The multi-factor recommendation
model is incremented with more factors, including time decay, tagging activity and
location. Location is also improved with a method to predict event trajectories. Moreover, a
new family of tensor-based recommenders is presented. The tag cloud model is improved with
two new tags selection algorithms. We also introduce a propagation of the (ir)relevant terms
obtained from tag clouds usage to WP3 and WP4 in order to improve their data collections
algorithms. The evaluation part of the report summarizes the benefits and drawbacks of the
models presented and describes possible directions for the future work.
Original language | English |
---|
Number of pages | 68 |
---|---|
Publication status | Published - Aug 2012 |
Keywords
- social tagging
- recommender system
- tag cloud
- group recommendation
Fingerprint
Dive into the research topics of 'M-Eco enhanced Adaptation Service (D5.3)'. Together they form a unique fingerprint.Projects
- 1 Finished
-
MEco: MEco - Medical Ecosystem - Personalized Event-Based Surveillance
Dolog, P., Xu, G., Lage, R. G., Bayyapu, K. R. & Pan, R.
01/01/2010 → 30/06/2012
Project: Research