IoT based air components collection for machine learning reinforcement

Abstract

Air pollution has been a noteworthy threat for a long time now in the 21st century. Human lives have never faced such an obscene amount of threat from the very air it needs to breathe to stay alive. As technology evolves more and more with every passing month, year, and decade, the emissions caused by the modern utilities are increasing as well. The measurement of air quality is done through an index called “AQI” which elaborates as the Air Quality index. The proposed work revolves around the collection of air component data through an IoT device and determining the AQI periodically and creating a proper dataset for the air quality index of the city of Dhaka. The IoT device is configurable to receive sensor data periodically. MQ- 7, MQ-131, MQ-135 for air component detection, PMS5003 for particulate matter detection, DHT11 for humidity and temperature measurement and RTC DS3231 real-time clock module for timestamp has been used to make the device a complete frontrunner for a cheap data collection source. The data collection has been curated in such a way that pre-processing of datasets for certain machine learning and deep learning algorithm get much easier. All the sensors and modules are connected and worked in harmony by connecting them to a microcontroller (Arduino) and is stored and accessed remotely via an MPU (Raspberry Pi). The remote access is granted via cloud service (VNC Viewer). The acquired datasets are then ran through machine learning and deep learning layers (such as Random forest, Lasso Regression, Linear Regression, KNN, LSTM etc.) for the further prediction of the AQI.