Outlier detection using k-means clustering and lightweight methods for Wireless Sensor Networks
Aujor Tadeu C. Andrade, Carlos Barros Montez, Ricardo Moraes, Alex Sandro Roschildt Pinto, Francisco Vasques, Gabriel Lopes Da Silva · 2016
Wireless Sensor Networks (WSNs) arc susceptible to faults both in sensors and in communication. Information fusion techniques allow to extract precise information from a large amount of data. Detection, identification and treatment of outlier, in these techniques, is a key point. Outlier detection in WSNs is a challenge due to the low capacity of the nodes and low bandwidth of the network. This paper proposes a methodology that applies the clustering and lightweight statistics techniques for detection of outliers in WSNs. The assessment of the methodology involves a case study with temperature sensors in WSN nodes. The results show that this methodology is able to provide precise information, even in the presence of outliers.