System Design for Air Quality Monitoring and Classification using Artificial Neural Network with Hyperparameters Optimization

Abstract

Air pollution in both urban and rural areas was estimated to have died 4.2 million people worldwide, with developing countries accounting for 91% of these deaths. Indonesia, a developing country, is one of the world’s ten most polluted countries. AQMS (Air Quality Monitoring Station) which is static and only numbered 36 units throughout Indonesia make the process of monitoring air quality less evenly distributed to the regions. Therefore, in this research will be made a device to monitor and classify air quality by mobile. This study will focus on monitoring air quality with sensors and then displaying the results on a website to make it easier for users to access and read the results of air quality mapping, this website is GIS-based website. ULPSM sensors were used in this study for gas parameters CO, NO2, O3, and SO2, and PMS5003ST sensors for PM2.5 and PM10 parameters. The air quality classification method using ANN (Artificial Neural Network) with hyperparameter tunning approach for a better model performance. The best hyperparameter variables is confirmed trough the experiments resulting the better prediction accuracy.