This study develops a data-driven framework for predicting the thermal stability of crystal-doped polymer composites and evaluating multifactorial effects based on a structured materials database and machine learning methods. Feature variables related to the polymer matrix, crystalline dopants, composite formulation, processing conditions, and thermal analysis parameters were systematically organized, and predictive models were constructed for key thermal stability indicators, including T5%, T10%, Tmax, and char residue. The results indicate that ensemble learning models can more effectively describe the nonlinear relationships between thermal stability and multi-source variables than conventional linear models, with the XGBoost model exhibiting higher prediction accuracy and better generalization capability. Further feature importance analysis and multifactorial effect evaluation reveal that the intrinsic thermal stability of the polymer matrix, crystalline dopant type, filler loading, particle size, surface modification, testing atmosphere, and heating rate are key factors governing thermal stability. An appropriate amount of crystalline dopant can improve the thermal degradation behavior of polymer composites through physical barrier effects, interfacial restriction, heat-transfer regulation, and catalytic char formation, whereas excessive loading may weaken the enhancement efficiency due to filler aggregation and interfacial defects. The proposed predictive model and multifactorial evaluation strategy provide a data-driven basis for the rational design of crystal-doped polymer composites with enhanced thermal stability.
