Physically interpretable prediction of polymer dielectric constants for accelerated material screening in energy storage systems

• Simplifies dielectric constant prediction by using only repeating unit structures, avoiding complex calculations. • Uses easy-to-understand parameters (e.g., elemental composition, polar groups) for clear insights. • Tested on 114 experimental values, ensuring reliability for a wide range of polymers. • Beats existing models with better accuracy (higher r², lower errors like RMSE and MAE). • Provides a practical, user-friendly tool for researchers and industry professionals. This work introduces a simple and physically interpretable multiple linear regression model that predicts dielectric constants of organic polymers directly from repeating-unit elemental composition and acyclic amine content, without reliance on proprietary software or high-dimensional computer-generated descriptors. By using and extending established literature datasets, a curated benchmark of 114 polymers at 298 K and 100 Hz is assembled, achieving excellent accuracy (r² > 0.96, RMSE < 0.12) together with rigorous internal and external validation, including an independent 40-polymer external set (r² = 0.939). In contrast to previous QSPR and machine-learning approaches that depend on dozens of abstract descriptors and extensive hyperparameter tuning, the proposed model uses only four chemically transparent descriptors—elemental contributions, acyclic amine groups, and two structural correction factors—that directly connect atomic makeup and polar functional groups to macroscopic dielectric response.