Currently, in Republic of Korea, most of the research that requires huge computational resources utilizes the supercomputer resources of the National Supercomputing Center. However, the average utilization rate of the resources for the past three years has reached 80%. In order to secure the stability of the national infrastructure, the establishment of specialized supercomputing center infrastructure is being promoted to distribute computational demand by field. Therefore, it is necessary to accurately predict the computational demand distributed to the specialized centers in order to build an appropriate scale of infrastructure for the specialized centers in the future. In this paper, a preference survey was conducted for national center and specialized centers by using the results of individual behavior surveys through logit model analysis applying probability utility theory. Through the computational demand share of the national center and the specialized center derived through the analysis results, it will be used as the highest level of evidence when calculating the infrastructure size of the specialized center for each field in the future.