This analysis demonstrates improved CO2 prediction in a hospital setting, suggesting better IAQ management strategies. Attention-based LSTM shows enhanced forecasting accuracy with multiple predictors.
Given the significant implications of indoor air pollution for physical and mental health, well-being and productivity, indoor air quality (IAQ) is of critical importance. CO2 is a prevalent indoor air contaminant and represents a key determinant for IAQ characterization. This study collected sensed air pollution and climatic data from a hospital environment in Greece and employed Long Short-Term Memory (LSTM) neural network variants with progressively increased architectural complexity to predict indoor CO2 concentration across future horizons ranging from 15 min up to 180 min. Among the examined variants, the attention-based LSTM exhibited the most consistent performance across the forecasting horizons. Incorporating additional predictors reflecting climatic conditions, air pollution and occupancy status within the hospital settings, the multivariate attention-based LSTM further enhanced its predictive performance with an MAE of 8.9 ppm, 16.7 ppm, 31.2 ppm, 38.9 and 39.5 ppm for 15 min, 30 min, 60 min, 120 min, and 180 min ahead, respectively.
No takes yet. Share an insight, caveat, or question.
Mountzouris et al. (2025) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: