Towards safer indoor spaces: A machine learning based CO₂ forecasting approach for smart systems

Indoor Air Quality (IAQ) is a significant concern for public health, with indoor air pollution contributing to millions of deaths annually. This research proposes a machine learning (ML) based approach to forecast CO <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> levels indoors, a key indicator of IAQ. By utilizing timeseries forecasting and artificial intelligence (AI) techniques, the study aims to predict CO <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> concentrations in a short term horizon (five minutes) and enhance indoor air quality. Two ML methods, Long Short-Term Memory (LSTM) networks and dilated Convolutional Neural Networks (CNNs), were explored and compared to naive forecasting methods. The best performing model achieved a Mean Absolute Error (MAE) of 4.56 ppm. Results demonstrated the potential of ML-based approaches to accurately predict CO <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> levels, which can be utilized to optimize the use of Heating Ventilation and Air-Conditioning (HVAC) systems for safer indoor spaces.