Frequent radio transmissions among sensors, or from sensors to the basestation, have always been a major energy drain. One of the approaches to reduce the data transmitted to the basestation is to shift the bulk of data processing to networked sensor nodes; for instance, sensors to send only data aggregates to reduce the overall amount of data exchanged. Sensor nodes, however, are quite limited in terms of their energy and processing power, and as such, traditional centralised data mining algorithms are infeasible to be directly implemented on sensors. In this paper, we modify APRIORI to find strong rules from sensor readings in a sensor network and using these rules, autonomously control sensor network operations or supplement sensor operations with a rule knowledge base. For example, triggers activated from the rules could be used to sleep sensors or reduce data transmissions to conserve sensor energy. Our work here includes a detailed implementation of a lightweight rule learning algorithm for a resource-constrainted sensor network, with simulation results for a group node setup running the algorithm.
|Number of pages||5|
|Publication status||Published - 16 Mar 2008|
|Event||Proceedings of the 2008 ACM Symposium on Applied Computing - Fortaleza, Ceara, Brazil|
Duration: 16 Mar 2008 → 20 Mar 2008
|Conference||Proceedings of the 2008 ACM Symposium on Applied Computing|
|Period||16/03/08 → 20/03/08|