Multi-scenario assessment of living-wall effects on building air purification, noise attenuation, and thermal insulation using quantum machine learning

Despite a surge in the deployment of building-integrated Living Wall Systems (LWS) to foster urban sustainability, the complexity of field measurements has led to disjointed and often insufficient quantification of contributions to air purification, noise abatement, and thermal stability. This study combines multi-scenario ecological assessment with a dedicated Quantum Machine Learning (QML) model to precisely extract features for complex environmental mitigation. Laboratory-scale LWS unit boxes with and without artificial pollution sources, and an in-situ test on an occupied LWS-equipped building were selected. An IoT-based monitoring system was developed to collect internal and external PM₂.₅, PM₁₀, SO₂, NO₂, O₃, eTVOC, temperature, humidity, and noise with 1-minute intervals. A QML approach was developed, and accuracy was evaluated using R², RMSE, and ROC curves, compared with conventional random forest (RF) and artificial neural network (ANN) based models. A cross-scenario contribution analysis decomposed each indoor environmental variable into outdoor drivers to quantify the relative mitigation attributable to LWS. Under polluted conditions, SO₂ and NO₂ removal exceeded 60%, PM₁₀ interception was about 40%, aggregated average noise attenuation in the building reached 25.4 dB, and sun-exposed façades exhibited a consistent cooling effect. The QML model outperformed RF and ANN, with R² ranging from 0.82 to 0.99, compared with 0.71 to 0.96 for RF; 0.35 to 0.90 for ANNs, and lower RMSE. This study provides quantitative evidence for the use of LWS in green buildings and introduces a quantum approach that redefines optimization methods for comparable indoor environments.