The two main constraints of sensor networks are the high number of data sources and the sensors' limited ressources. These constraints call for specific data mining architectures to process information. In this talk, we propose a two-layer modular architecture to adaptively perform supervised learning tasks in large sensor networks. We show that a clustering procedure along with the combination of two algorithms, recursive PCA and lazy learning, are well suited to sensor network constraints.
Sensor networks, Supervised learning, Online PCA, Lazy learning
G. Bontempi, Y. Le Borgne. (2005)
An adaptive modular approach to the mining of sensor network data, .
Accepted for publication.