A Multi-Sensor Fusion Framework Based on Machine Learning for COPD Pulmonary Function Prediction

Continuous and easily accessible pulmonary function testing is desperately needed for chronic obstructive pulmonary disease (COPD), yet traditional clinical tests are expensive and inappropriate for home-based management. In order to solve this, we provide a multi-sensor fusion framework based on machine learning that uses artificial intelligence (AI) to provide low-cost, real-time pulmonary function prediction. The system combines respiratory signals obtained by RFID-based RSSI measurements with physiological data (heart rate and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{SpO}_{2}$</tex>) from wearable sensors. Predictive modeling, SHAP-based feature selection, and data preparation all use advanced AI techniques. We improve the interpretability of the model while preserving predicted performance by using SHAP analysis to determine the top 12 contributing features. When it comes to learning complex nonlinear patterns from multimodal biosignals, XGBoost performs the best among the examined models (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{R} 2=0.967$</tex>). The potential of AI-enhanced health analytics to provide scalable, noninvasive, and intelligent pulmonary monitoring tools for COPD home rehabilitation is highlighted by this work.