Intelligent Wireless Sensor Network for Gas Classification Using Machine Learning

Abstract

Systems based on wireless sensor networks convey a prevailing tool to monitor data in complex environments. In this work, we present an intelligent integrated system that monitors various crucial environmental parameters, and applies machine learning to classify multiple gases. The system is a complete, sophisticated, interactive multiple parameters-observing, with two-way communication via wireless mobile network. It consists of a number of buoys distributed throughout the sea and integrated with a suite of sensors capable of measuring a range of essential air and underwater parameters. Four well-known machine learning algorithms are implemented and compared, namely multilayer perceptron, Naïve Bayes, logistic regression, and support vector machines. We present a thorough investigation and deep analysis to evaluate the system performance. The experimental results show that the system achieves a correct classification rate of 92.66%, with a weighted average precision, recall, F-score, and MMC of 0.93, 0.93, 0.926, and 0.91, respectively.