Voting Ensemble Learning Model (VELM) for Harmful Gas Detection Systems: A Proposed Model

Abstract

An increase in industrialization has caused the release of harmful gases into the atmosphere and this has resulted in environmental pollution that greatly affects the health of both plants and animals alike. Over time, sensor-based technology has been used to develop gas detection systems, but they were faced with some major challenges, such as sensor failure and poor performance that have greatly affected their performance. Also, the use of machine learning techniques has been developed for this purpose by leveraging the powers of sensor-based technology. However, they were faced with issues, such as poor feature selection criteria and missing data, which resulted in delays and inaccurate prediction performance of the models. This study, therefore, seeks to survey existing ensemble learning models and to design a Voting Ensemble Learning Model (VELM) for harmful gas detection systems. The methodology adopted for this study is a comparative analysis of the classification models used for the study. The study concludes that there is a need to develop advanced machine learning models to be used for detecting harmful gas and other detection systems such as earthquake, volcano, and landslide detection systems, in real-time that can inform future research. Also, a recommendation was made for the implementation of the developed model based on its low variance and low bias when dealing with widely dispersed data points in a dataset in addition to its ability to deal with local minima and overfitting that affect classic machine learning. Keywords: Detection System, Ensemble Learning, Machine Learning, Sensor