Data mining in multi-feature sensor networks

Wireless Sensor Networks have been used in many of critical applications such as battlefield surveillance and monitoring of critical infrastructures. Such applications may require heterogeneous sensor nodes with different sensed phenomena and different hardware characteristics. This heterogeneity imposes an added constraint to the mining of useful information from the network. At the same time, a stream of data that is frequently reported from each node to the sink node may be needed. Therefore, mining such online stream of data becomes a challenging task; especially in a network with limited capabilities such as sensor networks. Moreover, a sensor node might measure more than one phenomenon. We call such network a multi-feature sensor network. In this paper, we propose a global sensor network framework and we suggest different algorithms that can be used in each layer. However, our focus in this paper is on developing a new clustering algorithm that handles the concept of multi-feature sensor networks where each node reports more than one feature. For this purpose, we introduce Multi-Feature LEACH based Clustering algorithm (MFLC). In contrast to LEACH, MFLC considers the number of features reported by each node as well as the nodes´ residual energy during the clustering process. MFLC is evaluated through extensive set of experiments.