Aiming at reducing energy consumptions in data writing, the development of new memory materials is highly required. To develop new phase change materials (PCMs) with extremely low operation energy used in non-volatile memory, such as storage-class memory, we performed Bayesian optimization for the physical properties of PCMs through numerical simulations. In this numerical simulation, the electrical potential and temperature distribution were solved simultaneously. It was found that a PCM with low thermal conductivity, low melting temperature, and low ratio of contact resistance to volumetric resistance results in a low operation energy of a PCM-based memory application. Finally, we developed a design strategy for PCMs. New PCMs should be developed by lowering the operation energy E, described as E = κ(1 + C)ΔT/Δz, where κ is the thermal conductivity of the PCM, ΔT is the melting temperature, C is the ratio of contact resistance to the volumetric resistance, and Δz is the thickness of the PCM. The present results clarify the relationship between thermal and electrical properties for lowering operation energy ever hidden in previous studies. According to the design strategy, an operation energy in phase change memory application can be decreased to less than 1/100 compared with conventional Ge-Sb-Te compounds.