Article

Predicting Perceived Learning-Environment Quality in Classrooms Using Indoor Environmental Measurements and Machine Learning

---

Abstract

Indoor Environmental Quality (IEQ) plays a critical role in shaping students’ classroom experience, yet linking objective environmental measurements to subjective perceptions remains methodologically challenging. This study investigates whether perceived learning-environment quality can be predicted from indoor environmental measurements and contextual variables using supervised machine learning. A longitudinal dataset containing repeated satisfaction surveys and temporally aligned environmental measurements from classrooms in Belgium was used. The targets included thermal, indoor air quality (IAQ), acoustic, and visual satisfaction scores measured on ordinal Likert scales. To ensure robust evaluation and prevent information leakage, a group-based train–test split was applied at the occupant level, guaranteeing generalization to unseen individuals. Multiple models were evaluated, including a median baseline, Ridge regression, Random Forest, and HistGradientBoostingRegressor. Performance was assessed using Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and Quadratic Weighted Kappa (QWK) to capture ordinal agreement. All machine learning models significantly outperformed the baseline. Ensemble models achieved macro-average QWK values above 0.70, indicating strong ordinal consistency. IAQ and thermal satisfaction were most predictable, while acoustic and visual satisfaction showed greater variability. Most errors occurred between adjacent Likert categories, and extreme misclassifications were rare, demonstrating stable ordinal behavior. The results confirm that measurable environmental factors contain sufficient predictive signal to approximate subjective classroom satisfaction under realistic generalization conditions. Linear models captured much of the structure, while boosted trees provided marginal performance gains with improved flexibility. These findings support the feasibility of data-driven approaches for monitoring and improving classroom environments and provide a reproducible modeling framework for future research in educational building analytics.

Keywords: Indoor Environmental Quality, Classroom Comfort Prediction, Machine Learning, Ordinal Regression, Educational Building Analytics

How to Cite:

Mohammed, E., (2026) “Predicting Perceived Learning-Environment Quality in Classrooms Using Indoor Environmental Measurements and Machine Learning”, Artificial Intelligence in Learning 2(1). doi: https://doi.org//AIL.144