Machine Learning for Real-Time Environmental Monitoring and Air Quality Prediction
Acep Purqon(a*), Irfan Dwi Aditya(b), Berlian Oka Irvianto(c), Sparisoma Viridi(d)

a) Earth Physics and Complex System Research Division
b) Physics of Instrumentation and Computation Research Division
c) Master Program in Physics
d) Nucler Physics and Biophysics Research Division
a,b,d) Data Science in Physics Specialization
Physics Department, Institut Teknologi Bandung

Abstract

We are developing a method to analyze real-time environmental data using machine learning. The data is collected from sensors that measure carbon dioxide (CO&#8322-) concentration, temperature, humidity, wind speed and direction, and fine particle levels (PM2.5). The sensor data is transmitted using the public MQTT broker at mqtt://mqtt.eclipseprojects.io. Our goal is to detect patterns in the data, understand the relationships between environmental variables, and predict changes in air quality and climate conditions. We use several machine learning techniques, including regression, classification, and deep learning. Before training the models, we clean the data to handle noise and missing values, and select key features to improve model accuracy. Model validation is also performed to ensure reliable predictions. The analysis accompanied with visualization is expected to provide insights that can support environmental policy and help address climate change.

Keywords: Machine learning- environmental data- CO&#8322-- PM2.5- temperature- humidity- wind- MQTT- air quality prediction

Topic: Energy and Environmental Physics