Banking operational services are supported by a banking information system called Core Banking System (CBS). Due to frequent request from branches without accompanied by standard references, risk of improperly granting of access right occurs. Users have their accounts already codified into specific pattern according to one's role. However, it is still difficult in grouping user accounts into their proper roles as current access rights are greatly vary and situation exists where a particular user has multiple roles. Steps to solve those problems in this research are preparing data in order to process user access rights recorded in CBS, clustering analysis to obtain optimum grouping of access rights, and evaluation of clustering results. Our experimental results show better cluster produced, measured by correctly labelled roles. However the results also produces inconsistent sets of user menus associated with mined roles, due to significance of some features that represents branch office class and user position.