TinyMLAir: Indoor Air Quality of Domestic Activities through TinyML

Research has largely concentrated on outdoor air pollution, often neglecting the significant impact of indoor air pollution on daily health. Our research addresses this gap by investigating the effect of indoor activities regarding the pollution. We use 24-hour data to personalize assessments and pinpoint activities driving indoor pollution for targeted reduction strategies. We developed a comprehensive monitoring system with a Raspberry Pi Pico board, analyzing 3720 minutes of diverse data, including incense burning, indoor smoking, cooking in poorly ventilated areas, heavy air conditioner use, and accidental paper combustion. By integrating machine learning models, we achieved a 90.60% accuracy with Random Forest. We propose an innovative short-term pollution intake alert system to mitigate long-term health risks associated with indoor pollution, promoting individual well-being.