| Wireless sensor networks provide opportunity to gather large volume of data for applications spanning from scientific, industrial, to military fields. The most challenging issue in the development of wireless sensor network applications is energy constraint. Recent works of node clustering in wireless sensor networks often rely on energy-centric protocols focusing on energy efficiency in order to prolong network lifetime. However, they have ignored data quality, another important factor in protocol design. Ignoring data quality limits exiting protocols to simple statistical type of applications, such as finding the MIN, MAX, or AVG of the sensor readings. They are not applicable to applications where the goal is to collect data for off-line analysis.;The problem we are trying to solve is to design energy-efficient and high data quality node clustering protocols that utilize both system data and application-specific sensor data, and are applicable to a wide range of wireless sensor network applications. In this work, we develop data-centric node clustering protocols by techniques of data mining. The analysis and comparison of both energy-centric and data-centric node clustering protocols on real-world datasets show that by incorporating application-specific sensor data into node clustering is effective in both prolonging network lifetime and assuring data quality. |
