On the stability of context prediction
Immanuel König, Bernd Niklas Klein, Klaus David · 2013
Context prediction is a key technique for proactive environments adapting to user's needs. To prevent wrong predictions is one key factor to achieve a high user acceptance. A wrong prediction could be caused by faulty or disturbed sensor data. With the triumph of the Smartphone, a wide range of context sources has become ubiquitous. Often, context prediction approaches today do not utilize these multiple context sources to cope with faulty or disturbed sensor data. We propose and evaluate an approach that uses multiple context sources and exploits the correlations between context sources of one user to get a more fault tolerant prediction.