Intelligent prediction of heat capacity of polyethylene glycol polymer

Polyethylene glycol (PEG) exhibits tunable molar heat capacity, making it essential for thermal management and materials processing. Using an experimental databank of 528 observations, this study applies machine learning to develop predictive models based on temperature and molar mass. Data reliability was ensured through Monte Carlo-based outlier detection and correlation analysis, which confirmed temperature as the dominant predictor. Among the algorithms tested, Random Forest delivered the highest performance (R² = 0.9969, AARE = 8.29%), outperforming other methods that showed signs of overfitting. The results demonstrate that Random Forest provides a cost-effective, accurate, and scalable alternative to calorimetric experiments, enabling precise estimation of PEG heat capacity for applications in materials design, energy storage, and biomedical engineering.