Soft sensor for predicting indoor PM_2.5 concentration in subway with adaptive boosting deep learning model.

Area of Science:

• Environmental Science
• Data Science
• Public Health

Background:

• Indoor air quality (IAQ) is crucial for public health.
• Accurate monitoring of indoor particulate matter (PM2.5) in subway environments is challenging.
• Soft measurement techniques offer a promising solution for IAQ monitoring.

Purpose of the Study:

• To develop and evaluate a novel soft sensor model for precise PM2.5 monitoring in subway environments.
• To enhance the accuracy and reliability of IAQ data analysis using ensemble learning.
• To improve the interpretability of the soft sensor model.

Main Methods:

• Kernel Principal Component Analysis (KPCA) was employed for nonlinear dimensionality reduction and noise reduction.
• Adaptive Boosting (AdaBoost) was utilized as an ensemble learning technique to enhance Long Short-Term Memory (LSTM) model performance.
• The proposed KPCA-AdaBoost-LSTM model was developed and validated for IAQ monitoring.

Main Results:

• The KPCA-AdaBoost-LSTM model demonstrated high predictive performance for PM2.5 concentrations.
• Achieved R-squared values of 0.9007 in the hall and 0.8995 on the platform.
• SHapley Additive exPlanations (SHAP) analysis provided insights into model interpretability.

Conclusions:

• The KPCA-AdaBoost-LSTM soft sensor offers a reliable and accurate method for monitoring subway IAQ.
• The model's enhanced performance and interpretability contribute to better public health strategies.
• This approach advances the application of machine learning in environmental monitoring.