Predictive Analytics of Sensor Data Based on Supervised Machine Learning Algorithms

Abstract

Wireless sensor network (WSN) is one of emerging technologies in today's scenario. Due to progressive advancement in micro-electro-mechanical system (MEMS) technology it can easily deployed in harsh environment. Sensor node communicates with their neighboring sensor nodes via radio frequencies and has many notable capabilities like self-configurable, self-manageable and monitoring physical phenomenon. Wireless sensor network is gaining popularity due to presence of many characteristics like cheap, cost-effective, reliable etc. along with this it has one major challenge that is limited battery life. To overcome this challenge, many solutions have found till date such as improvising routing protocols, reduction in computation of data, modification in time-stamp synchronization etc but still need more work. In this paper, our major focus is on processing of sensor dataset using various machine learning algorithms. We have managed different range of datasets from hundreds to thousands values and processed with various supervised machine learning algorithms. Simulation result shows that Gaussian Naive Bayes algorithm prominently gives better results than other algorithms in terms of accuracy parameter.