DEMS: A Data Mining Based Technique to Handle Missing Data in Mobile Sensor Network Applications

In Mobile Sensor Network (MSN) applications, sensors move to increase the area of coverage and/or to compensate for the failure of other sensors. In such applications, loss or corruption of sensor data, known as the missing sensor data phenomenon, occurs due to various reasons, such as power outage, network interference, and sensor mobility. A desirable way to address this issue is to develop a technique that can effectively and efficiently estimate the values of the missing sensor data in order to provide timely response to queries that need to access the missing data. There exists work that aims at achieving such a goal for applications in static sensor networks (SSNs), but little research has been done for those in MSNs, which are more complex than SSNs due to the mobility of mobile sensors. In this paper, we propose a novel data mining based technique, called Data Estimation for Mobile Sensors (DEMS), to handle missing data in MSN applications. DEMS mines the spatial and temporal relationships among mobile sensors with the help of virtual static sensors. DEMS converts mobile sensor readings into virtual static sensor readings and applies the discovered relationships on virtual static sensor readings to estimate the values of the missing sensor data. We also present the experimental results using both real life and synthetic datasets to demonstrate the efficacy of DEMS in terms of data estimation accuracy.

In DMSN' 2010