Data Fault Detection in Wireless Sensor Networks Using Machine Learning Techniques

Wireless Sensor Network (WSN) is a wireless network that consists of spatially distributed autonomous devices with sophisticated subsystem called sensors to monitor the environmental conditions. WSN’s measure environmental conditions like temperature, sound, pollution levels, humidity, wind, etc. Data fault detection is a challenging problem due to presence of sensors in unforeseeable areas. This detection has to be precise and accurate, so that it can be used for weather prediction, disease prediction, traffic monitoring, etc. In recent years, Machine Learning plays a vital role in accurate fault detection. In this paper, we propose an Enhanced Minimum Redundancy Maximum Relevance algorithm. Our work combines the principle of filter and wrapper methods by using F-Mutual Information Difference (FMID) and FMIQ as the objective functions to determine the relevance and redundancy. The core analysis carried out in various datasets shows that the classification accuracy ranges from 93 to 100%, which is promising compared to the existing systems. The strength of our algorithm can be leveraged in situations where data fault detection accuracy is a real matter of concern.