Predictive Analytics for Carbon Dioxide Levels A Linear Regression Approach

Abstract

Air quality in classrooms is very important for health and comfort. Increased concentrations of
carbon dioxide (CO2) due to human activities can cause health problems such as headaches, fatigue
and respiratory problems. This research develops an Internet of Things (IoT) based air quality
monitoring system using the HC-SR04 sensor to count the number of people and the MQ-135
sensor to measure CO2 levels. Data is integrated with the Thinger.io platform in realtime and
classified based on IQ Air standards. The IQ Air classification consists of : >1600 PPM in the
good category, <=2500 PPM in the moderate category, <=3300 PPM in the unhealthy category
for sensitive groups, <=4100 PPM in the unhealthy category, <=4900 PPM in the very unhealthy
category, and <=5500 PPM in the hazardous category. The LED indicator and buzzer provide
visual and audio warnings based on detected CO2 levels. The linear regression method was used
to calibrate the MQ-135 sensor, showing a high level of accuracy with an error percentage of
1.9%. The test results show that this system provides accurate and realtime information to
monitor indoor air quality, provides early warning if CO2 levels reach unhealthy levels, and can
improve air quality in classrooms.