Compared to HVAC, the application experience of composite/polymer insulators in HVDC transmission lines is limited. However, the tendency to utilize them in HVDC transmission lines, where pollution flashover is a major problem, is growing due to their water-repellent feature. This trend demands understanding insulator flashover performance and recognizing key parameters in the design and dimensioning of insulators. This paper presents the measurement and analytical results for dc pollution flashover of silicon rubber insulators at extra heavy pollution conditions. Two approaches are employed to model flashover voltage gradient and quantify the contribution of salt deposit density, specific leakage distance, average diameter, shed spacing to shed depth ratio, and form factor to insulator pollution performance. Developing a nonlinear regression model, the effect of the mentioned parameters on pollution flashover is studied. Then, the artificial neural network (ANN) and randomization method are used to develop a model to interpret parameters contribution to pollution flashover, and the results are validated with those of the regression method. Finally, by applying an ANN-based randomization approach, a multi-input model is created to quantify the significance of different parameters in flashover voltage gradient. The model ranks specific leakage distance and average diameter, respectively, as the most and least important geometrical characteristics in pollution flashover performance of SiR insulators. The method and results can be a source of information to optimize the design and dimensioning of HVDC SiR insulators especially in severe pollution conditions.