Skip to Main Content
Estimating missing sensor values is an inherent problem in sensor network applications; however, existing data estimation approaches do not apply well to the context of datastreams, a major characteristic of sensornet applications. Additionally, they fail to account for relationships among sensors and simultaneously, incorporate the time factor making the estimation process computationally aware of the relative relevance of each data round in the datastream. To address this gap, we propose a data estimation technique, FARM, which uses association rule mining to discover intrinsic relationships among sensors and incorporate them into the data estimation while taking data freshness into consideration. FARM was tested with data from two real sensornet applications, namely climate sensing and traffic monitoring. Simulation shows that in terms of estimation accuracy, FARM outperformed existing techniques costing only marginally more space and time overheads while scaling well with the network size, thus assuring quality of service for real-time applications.