Wireless sensor network fault detection via semi-supervised local kernel density estimation

Wireless sensor network (WSN) has become widely used in different applications. Fault detection of sensors is importance for maintaining a reliable WSN operation. And identification of faulty nodes in a WSN can be transformed into a pattern classification problem. In this paper, we introduce an effective label propagation procedure using semi-supervised local kernel density estimation. The proposed method estimates the posterior probability of a scene belonging to the faulty and it can preserve the manifold structure of dataset due to the utilization of kNN kernel for density estimation. Simulations based on a WSN are presented to show the effectiveness of the methods. The results demonstrate that our proposed algorithm can achieve better classification performance compared with other state-of-art semi-supervised learning methods.