Indoor air quality (IAQ) in academic environments has gained critical importance due to its direct implications on student cognitive performance, health, and well-being. Small academic rooms, characterized by limited floor area, high occupant density, and constrained architectural configurations, present unique challenges for maintaining adequate air quality through natural ventilation alone. This study presents a comprehensive evaluation of natural ventilation influence on indoor air quality parameters within small academic rooms at an Indian university campus, employing a custom-designed Internet of Things (IoT)-based real-time monitoring system.

A network of low-cost, calibrated multi-parameter sensor nodes was deployed across twelve classrooms of varying orientations, window-to-floor-area ratios (WFR: 0.08 to 0.22), and occupancy levels (15 to 45 students). Each node continuously measured carbon dioxide (CO2, ppm), particulate matter (PM2.5 and PM10, ug/m3), temperature (degC), relative humidity (%), volatile organic compounds (VOC, ppb), and formaldehyde (HCHO, ug/m3) at one-minute intervals over a monitoring period of nine months spanning all three climatic seasons (winter, summer, and monsoon).

Results indicate that CO2 concentrations exceeded the ASHRAE 62.1 threshold of 1000 ppm during 68.4% of occupied hours in rooms with WFR below 0.12 and adverse wind conditions. PM2.5 levels surpassed the WHO 24-hour guideline of 15 ug/m3 in 54.2% of monitored sessions during the pre-monsoon season, attributable to outdoor particulate infiltration through open windows. Rooms with cross-ventilation configurations demonstrated 34.7% lower mean CO2 concentrations compared to single-sided ventilation arrangements. Occupancy density emerged as the strongest predictor of CO2 accumulation (R2 = 0.87), followed by wind speed (R2 = 0.73) and window opening area (R2 = 0.68). Regression models and machine learning classifiers (Random Forest, AUC = 0.94) were developed to predict IAQ category (Good/Moderate/Poor) using environmental and architectural input parameters.

The IoT system demonstrated a mean absolute error (MAE) of 28.4 ppm for CO2 prediction and 1.8 ug/m3 for PM2.5 estimation after field calibration against reference instruments. Findings highlight critical deficiencies in current natural ventilation practices and provide evidence-based recommendations for window design, occupancy scheduling, and supplementary mechanical ventilation strategies. This research contributes directly to sustainable building design and healthy learning environment policy in tropical climate zones.