Practical statistical models were developed to quantify individual contributions from characteristics of conventional and non-conventional fillers and predict resulting mechanical properties of both hevea and guayule natural rubber composites. Carbon black N330 and four different agro-industrial residues, namely, eggshells, carbon fly ash, processing tomato peels, and guayule bagasse, were used in this study. Filler characteristics were used as explanatory variables in multiple linear regression analyses. Principal component analysis was used to evaluate correlations among explanatory variables based on their correlation matrices and to transform them into a new set of independent variables, which were then used to generate reliable regression models. Surface area, dispersive component of surface energy, carbon black, and waste-derived filler loading were found to have almost equal importance in the prediction of composite properties. However, models developed for ultimate elongation poorly explained variability, indicating the dependence of this property on other variables. Agro-industrial residues could potentially serve as more sustainable fillers for polymer composites than conventional fillers. This new modeling approach for polymer composites allows the performance of a wide range of different waste-derived fillers to be predicted with minimum laboratory work, facilitating the optimization of compound recipes to address specific product requirements.