We proudly announce a new achievement accomplished by the University of Patras and Thriassio Hospital. A new journal publication has been published within the context of TwinAIR project!
The publication titled “Short-term forecast of indoor CO2 using Gradient Boosting: A use case of a hospital in Greece”, authored by Christos Mountzouris; Grigorios Protopsaltis; John Gialelis; Despoina Fytili and its respective DOI is the following: 10.1109/ETFA65518.2025.11205796
The abstract can be found below:
Abstract
Considering the significant implications of indoor air quality on health, productivity, comfort and well-being, accurate forecasting of indoor air contaminants is of the utmost importance for devising effective strategies and proactive control of ventilation in buildings. CO2 is a common air pollutant in indoor settings, with human exhalation serving as its primary source. This study investigates the potential of machine learning models—particularly Gradient Boosting—for short-term indoor CO2 forecasting using an extensive dataset of air quality measurements collected from a hospital in Greece. The predictive performance of Gradient Boosting was evaluated against a baseline model, which interpolates the latest observed CO2 levels, as well as against linear models that capture recent temporal patterns in CO2 concentrations, over forecasting horizons of 5, 10, and 15 minutes. The Gradient Boosting model significantly outperformed the baseline, achieving a 44.53% reduction in Root Mean Squared Error (RMSE) and a corresponding decrease in Mean Absolute Error (MAE). In contrast, both the baseline and the linear models exhibited relatively low errors for 5- and 10-minute forecasts, suggesting that their low complexity and minimal computational demands make them reliable, lightweight alternatives for short-term CO2 prediction. Moreover, the findings underscore the value of incorporating occupancy factors in forecasting indoor CO2 levels, thereby paving the way for improved data-driven indoor environmental management strategies.
You can reach the full article following this link: https://ieeexplore.ieee.org/document/11205796