EK-means: A new clustering approach for datasets classification in sensor networks

Abstract

In wireless sensor networks (WSNs), hundreds or thousands of nodes are deployed in order to provide high quality monitoring. Nowadays, they constitute one of the most important sources of big data. Such amount of collected data is a real challenge for sensor nodes suffering from many limitations, especially, energy constraints. Therefore, to address big data issue, research efforts have been done today to design efficient data management (acquisition, aggregation, mining, etc.) techniques for WSNs. The main objective of these works is to reduce the amount of transmitted data over the network while preserving their properties. In this paper, we present a new data handling approach in order to reduce data transmission without the loss of data integrity. Our proposed approach, named EK-means, is a two-steps approach. First, it eliminates similar data generated at the sensors level using an Euclidean distance based data aggregation technique. Second, it applies an enhanced k-means clustering algorithm in order to group similar datasets generated by neighboring nodes into same clusters and reduce further the amount of data sent to the sink. Experiments on real sensor data show that our proposal can effectively minimize the energy consumption in WSNs and largely outperform the classic K-means algorithm.