Applying explainable artificial intelligence and ensemble boosting based machine learning models in predicting the thermal conductivity of nano enhanced phase change materials

This study attempts the integration of explainable artificial intelligence and ensemble boosting based machine learning framework (XAI-EBBML-F) in predicting the thermal conductivity of nano enhanced phase change materials (NePCMs). The XAI-EBBML-F includes the following boosting based models: light gradient (LightGBM), extreme gradient (XGBoost), natural gradient (NGBoost) and categorical (CATBoost). For this, historical datasets (n = 510) are inferred from the selected three classifications of literature which specifically dealt with single-walled carbon nanotube (SWCNT), multi-walled carbon nanotube (MWCNT) and carbon nanofibers (CNF). From the inferred datasets, 70% of the datasets were used for training and 30% for testing purpose. For the supplied datasets, the NGBoost model in the XAI-EBBML-F recorded the best prediction performance (with R² = 0.9820, RMSE = 0.0692, and MAE = 0.0301 for training and R² = 0.9677, RMSE = 0.1208, and MAE = 0.0533 for testing). Also, the results of interpretability analysis identified concentration of NePCMs as the most influencing factor in deciding its thermal conductivity.