A new in-network data reduction mechanism to gather data for mining wireless sensor networks

Abstract

Recently, association rules for sensors have received a great deal of attention due to their importance in capturing the temporal relations between sensor nodes in wireless sensor networks (WSNs). Because of this capability, these rules can be used to improve the Quality of Service (QoS) of wireless sensor networks by participating in the resource management process. To mine sensor association rules, behavioral data that describes the sensors' activities over time must be extracted and accumulated at the central node (the Sink) for further analysis. Given the limited resources of sensor nodes, a well designed data gathering algorithm is required for gathering the behavioral data efficiently. In this paper, an in-network data reduction technique is proposed to reduce the amount of data that needs to be routed to the Sink by exploiting the redundancy between sensors' activities. The in-network reduction technique will be implemented on top of a data gathering tree that we refer to as the Minimum Nodes Data Gathering Tree (MNDGT), which consists of the nodes that will participate in formulating the sensor rules and those that are necessary for maintaining the minimum distance to the Sink. To report on the performance of the reduction mechanism, a comparison analysis with other two gathering schemas is introduced. Indeed, the results show that the in-network reduction technique is able to reduce the number of messages by a factor ranging from 10% to 70% compared to the other techniques.

Anomaly detection in monitoring sensor data for preventive maintenance

“…Recently, the problem of mining sensor data has been addressed by the data mining community. Different approaches focusing either on the data representation by performing for example sensor clustering Ci, Guizani, and Sharif (2007), Rodrigues and Gama (2006), or knowledge extraction by mining association rules Boukerche and Samarah (2007), Boukerche and Samarah (2008), Chong et al (2008), Ma et al (2004), Mihail Halatchev (2005, Yairi, Kato, and Hori (2001), or sequential patterns Cook et al (2003), Guralnik and Haigh (2002), Tseng andLu (2009), Wu, Peng, andChen (2001) were proposed.…”