A Voting Ensemble Technique for Gas Classification

Abstract

This article discusses the factors that influence gas classification results such as the data pre-processing and the type of classifier, these are two important factors in the electronic nose algorithm. Early in the data pre-processing process, machine learning algorithms are predominantly used for classification of the gas data, such as K-Nearest Neighbor (k-NN) and Support Vector Machine (SVM). A number of studies have been conducted throughout the past few years concerning the use of machine learning and neural network for gas classification. The focus of this paper is on gas classification and identification by using individual machine learning (Logistic Regression (LR), Naïve Bayes (NB)s, K-Nearest Neighbours (k-NN), Support Vector Machine (SVM), Decision Tree (DT), Random Forest (RF)) and ensemble (Stacking and Voting) techniques. Six different gases and a 4 × 4 sensor array is used for data collection. Using data collected by sensors arrays, it has been proven that our system is more accurate than individual classifiers. An improved accuracy of 98.04% is achieved by using Voting Classifier.