ASSESSMENT AND PERFORMANCE ANALYSIS OF MACHINE LEARNING TECHNIQUES FOR GAS SENSING E-NOSE SYSTEMS

Abstract

E-noses that combine machine learning and gas sensor arrays (GSAs) are widely used for the detection and identification of various gases. GSAs produce signals that provide vital information about the exposed gases for the machine learning algorithms, rendering them indispensable within the smart-gas sensing arena. In this work, we present a detailed assessment of several machine learning techniques employed for the detection of gases and estimation of their concentrations. The modeling and predictive analysis conducted in this paper are based on kNN, ANN, Decision Trees, Random Forests, SVM and other ensembling-based techniques. Predictive models are implemented and tested on three different MoX gas sensor-based experimental datasets as reported in the literature. The assessment includes a delineated analysis of the different models’ performance followed by a detailed comparison against results found in the literature. It highlights factors that play a pivotal role in machine learning for gas sensing and sheds light on the predictive capability of different machine learning approaches applied on experimental GSA datasets.