In this study, we investigate the melting behavior of ice at ice–polymer interfaces using molecular dynamics (MD) simulations combined with machine learning techniques. It is well established that even below the melting point, the surface structure of ice becomes disordered, forming a quasi-liquid premelting layer. This layer significantly affects interfacial phenomena such as friction, as exemplified by ice skating. Moreover, the thickness of the premelting layer depends on the material in contact with the ice; however, the molecular-level behavior of water in this region remains poorly understood. Therefore, we perform MD simulations to examine how hydrophilic and hydrophobic polymers influence the formation and characteristics of the premelting layer. Furthermore, we apply machine learning models to classify the phase (solid or liquid) of individual water molecules based on their local dynamical features. This approach enables a more detailed understanding of interfacial water behavior and the microscopic mechanisms underlying ice premelting in polymeric environments.