Classified Volatile Organic Compound Detection using Data Classification Algorithms

Abstract

Sensors are becoming smaller and less expensive, sparking interest in assessing vast volumes of sensor data. Meanwhile, the emergence of machine learning has led to the development of technologies that have a substantial impact on our lives. Machine learning models are often used to produce accurate, real-time predictions even in the presence of noisy sensed data. In this study, a Volatile Organic Compound (VOC) categorization system based on sensor data collected from a sensor array was developed. The most difficult challenge posed in the sensor array was the detection of the type of VOC. It is feasible to categorize VOCs brought on by applying data classification algorithms to data collected from sensor devices. In this work, we used data from the classification algorithms Decision Tree (DT), Naive Bayes (NB), and Linear Regression (LR) on a developed linear sensor array and their classification accuracy was compared. Four different VOCs were evaluated: acetone (C3H6O), benzene (C6H6), ethanol (C2H5OH), and toluene (C6H5CH3). The acquired classification accuracy reached 95.65% with the LR algorithm.