In the past few years, many wireless sensor networks had been deployed in the real world to collect large amounts of raw sensed data. However, the key challenge is to extract high-level knowledge from such raw data. In the applications of sensor networks, outlier/anomaly detection has been paid more and more attention. Outlier detection can be used to filter noisy data, find faulty nodes, and discover interesting events. In this paper we propose a novel in-network knowledge discovery approach that provides outlier detection and data clustering simultaneously. Our approach is capable to distinguish between an error due to faulty sensor and an error due to an event (probably an environmental event) which characterize the spatial and temporal correlations between events observed by sensor nodes in a confined network neighborhood. Experiments on both synthetic and real datasets show that the proposed algorithm outperforms other techniques in both effectiveness and efficiency.
展开▼
