Comparative Analysis of Data Mining Techniques Applied to Wireless Sensor Network Data for Fire Detection

Abstract

Wireless sensor networks (WSN) are a rapidly growing area for research and commercial development with very wide range of applications. Using WSNs many critical events like fire can be detected earlier to prevent loosing human lives and heavy structural damages. Integration of soft computing techniques on sensor nodes, like fuzzy logic, neural networks and data mining, can significantly lead to improvements of critical events detection possibility. Using data mining techniques in process of patterns discovery in large data sets it’s not often so easy. A several algorithms must be applied to application before a suitable algorithm for selected data types can be found. Therefore, the selection of a correct data mining algorithm depends on not only the goal of an application, but also on the compatibility of the data set. This paper focuses on comparative analysis of various data mining techniques and algorithms and in that purpose three different experiments on WSN fire detection data are proposed and performed. The primary goal was to see which of them has the best classification accuracy of fuzzy logic generated data and is the most appropriate for a particular application of fire detection.