Proton conductivity (σ) and hydration number (λ) are important characteristics for proton exchange membranes (PEMs) in fuel cells and water electrolyzers. A High σ yields high performance of these systems, while a low λ improves the dimensional stability to achieve the long-term operation. Accordingly, σ/λ is the most important characteristic because it reflects how efficiently protons can conduct at low water content and is called conduction efficiency in this study. For graft-type PEMs based on various types of polymers, σ, λ, and σ/λ depend on the base polymers, even at the same volume-normalized ion exchange capacity. We analyzed these three characteristics as objective variables using machine learning methods: least absolute shrinkage and selection operator (LASSO), artificial neural network (ANN), and random forest (RF) regression. A total of 20 physical and chemical properties of base polymers collected by literature, experiments, quantum chemical calculation, and descriptor generators were used as explanatory variables. The prediction accuracy was sufficiently high in the ANN and RF regression analyses, and low in LASSO. Only the RF regression model revealed the importance of explanatory variables. It is clear from the RF model that the enhancement of σ/λ should be achieved using base polymers with low crystallinity and high melting point.