Abstract:

Finding outlier data is very important for sensor network. In this paper we put forward the concept of local outlier by an example. Different with traditional outlier, whether a sensor node becomes a local outlier only depends on its neighbor nodes. We give its formula definition and put forward a Nave algorithm that consume excessive energy. Then we devise a distributed algorithm named DLODA. The DLODA divides the network into many areas and selects a cluster head for each area. The cluster head can compute part results innetwork and filter the data which will not become the results to save energy. Finally we evaluate the DLODA in terms of energy consumption and response time. Experiments show that the DLODA has good performance.

Key words: sensor network;local outlier detection;innetwork processing