Almost 400 amorphous polymer materials used in membrane gas separation technology are clustered on the basis of the shape of their polymer chains conformations. Obtained clusters, which rely solely on the geometry of polymer chains and not on the chemical class (polyamides, polyacetylenes, etc.), are shown to discriminate polymers with respect to their transport properties, in particular, the coefficient of diffusion. The method proposed consists of several steps. Firstly, realistic conformations of large polymer macromolecules are constructed using the program code developed in the RDKit environment for Python. Then, polymer conformations are characterized by a curve that relates the “accessible surface area” (i.e., the contact surface between the spherical model of a macromolecule and a spherical “probe”) to the radius of this probe, and also seven similar curves, which relate the polarized (neutral, positively or negatively charged, etc.) accessible surface area to the radius of the spherical probe that represents the variety of penetrant gases. An improved algorithm for surface area calculation maps out the outer surface of the macromolecule to eliminate its influence. The curves are averaged between ten polymer conformations to obtain more robust figures. Finally, agglomerative clustering is used to separate different polymers in the space of these curves that align their accessible-surface-area-related quantities against the probe radius. The proposed classification of polymers can be used to develop more precise predictive models of polymers’ transport properties for the theory-guided and computer-aided materials design.