Indoor Air Quality Prediction in Sick Building Using Machine and Deep Learning: Comparative Analysis

Abstract

Air pollution is a significant global concern that is continually increasing and threatening both the environment and human health. Air pollution is the principal factor leading to the deterioration of Indoor Air Quality (IAQ) in buildings. Carbon dioxide (CO₂) significantly contributes to indoor pollution intensifying, primarily from human activities. The demand for effective IAQ systems has increased due to the necessity for sustainable building development. The artificial intelligence (AI) models presented in this work utilized Machine Learning (ML) and Deep Learning (DL) methodologies to train the available dataset. This dataset was collected by the indoor sensors in Shanghai from November 2016 to March 2017 to predict CO₂ concentration and obtain pertinent information. The accuracy and the result of Ml and DL algorithms may differ depending on the datasets used and the algorithms' suitability for the specific data and application domain. Therefore, a significant benefit would be achieved by finding the best-fitted ML and DL approaches concerning the actual datasets and the application area. This necessity was fulfilled through an intensive review of the already existing DL and ML techniques. The DL techniques implemented in this analysis include Deep Adaptive Quantization Feature Fusion (DAQFF), Long Short-Term Memory (LSTM), Convolutional Neural Network (CNN), Gated Recurrent Unit (GRU), Bidirectional Long Short-Term Memory (Bi-LSTM), Random Forest-Long Short-Term Memory (RF-LSTM), CNN-LSTM, and Deep Neural Network (DNN). In contrast, the ML techniques include Decision Tree Regressor (DTR), Random Forest (RF), K-Nearest Neighbors (K-NN), Support Vector Regressor (SVR), and Gradient Boosting Regressor (GBR).