Context prediction is closely related to the learning of typical user context patterns. A decrease in the learning accuracy likely reduces the context prediction quality. We study the impact of context pre-processing on the learning efficiency of context prediction schemes. In particular, influences of the context data abstraction level on the learning accuracy are analysed. This discussion is accompanied by simulations as on sampled as well as synthetic context data. In the simulation an alignment prediction algorithm is applied to low-level GPS data and to high-level location data.
展开▼
