A Novel Data Mining Technique for Extracting Events and Inter Knowledge based Information from Wireless Sensor Networks

Advances in wireless sensor networks (WSNs) led to the emergence of a large number of applications for monitoring physical and critical environments. WSNs monitor the environments, detect events and report them to the applications. WSNs suffer from limited resources and yet they are responsible for delivering large amounts of data. This paper discusses the possibility of deriving useful information from reported data to gain insight about the status of the WSN, like predicting the existence of faulty nodes. A data mining technique is proposed to extract what we call chronological patterns. Chronological patterns can be thought of as tutorials that teach about the set of sensors that report events within a defined time interval. The proposed technique uses a new representation structure for the data extracted from WSNs. This structure, which we refer to as a chronological tree (CT), provides an efficient way of sub-pattern checking and a compressed format of large amounts of data. First, we introduce the way of extracting data from WSNs. We then discuss the CT structure and our proposed mining technique