Sliding window-based regularly frequent patterns mining over sensor data streams

Due to technical advancement WSNs are used in many applications that generate many stream nature data. An effective method to extract information from such streams is to apply associated rule mining. Association rule mining techniques use support metric (occurrence frequency) of a pattern as a criterion. In WSN data the patterns which maintain temporal regularity in their occurrence can reveal important knowledge. In this paper, we propose a technique to mine such patterns, called regularly frequent sensor patterns, employing a sliding window. Our Proposed technique uses a tree structure called regular frequent sensor pattern stream tree (RFSPS-tree) to store sensor data and then used an FP-growth like pattern-growth patterns to find regular frequent sensor patterns (RFSPs). Results show that the technique is efficient in finding RFSPs in terms of memory usage and computation time.