Related Experiment Video
Updated: May 14, 2026
07:12
Façade-Level Monitoring of CO2 Variability under Urban Heat Island Conditions using Low-Cost Sensor Data Loggers
Published on: December 12, 2025
Machine Learning Applications with Sensors for Indoor Air Quality Research
Cosmina-Mihaela Rosca1, Adrian Stancu2
1Department of Automatic Control, Computers, and Electronics, Faculty of Mechanical and Electrical Engineering, Petroleum-Gas University of Ploiesti, 39 Bucharest Avenue, 100680 Ploiesti, Romania.
Sensors (Basel, Switzerland)
|May 13, 2026
Summary
Indoor air quality (IAQ) research is vital as people spend most time indoors. This study reviews machine learning (ML) datasets and algorithms for IAQ, finding no single best model, emphasizing data quality for intelligent monitoring systems.
Area of Science:
- Environmental Science
- Computer Science
- Data Science
Background:
- Over 80% of daily life is spent indoors, making indoor air quality (IAQ) a critical factor for health and well-being.
- Machine learning (ML) offers promising solutions for monitoring and predicting IAQ parameters, but requires suitable datasets and algorithms.
Purpose of the Study:
- To provide a structured overview and comparative analysis of publicly available IAQ datasets for ML research.
- To map ML tasks and algorithms to IAQ prediction targets based on dataset characteristics.
- To investigate IAQ-ML using custom solutions and analyze trends in ML-based IAQ research.
Main Methods:
- Systematic literature review of 1162 (Web of Science), 1536 (Scopus), and 756 (IEEE Xplore) papers published between January 2020 and December 2025.
- Analysis of popular ML algorithms (Linear Regression, Logistic Regression, Random Forest, LSTM, PCA, Elastic Net) and their performance metrics (accuracy, R²).
- Investigation of custom IAQ monitoring solutions involving sensor data acquisition.
Main Results:
- Linear Regression, Logistic Regression, Random Forest (RF), and Long Short-Term Memory (LSTM) are the most frequently used ML algorithms for IAQ.
- High accuracies (>90%) and R² values (82-98%) are reported, with hybrid RF-LSTM achieving up to 99% R² for CO₂ and PM₂.₅ prediction.
- No universal ML algorithm exists for IAQ; data quality and structure significantly impact model performance.
Conclusions:
- Public and private IAQ datasets can support transfer learning, but require careful preprocessing for consistent results.
- The study provides a foundation for developing intelligent IAQ monitoring systems by highlighting effective ML approaches and the importance of data characteristics.
