In order to accurately grasp the satisfaction of users with different characteristics for electronic resources and effectively meet user resource needs, lay a solid foundation for the library's transition from "powerful resources" to "powerful connotative services" , the paper uses the method of questionnaire, RFD models, random forests from the perspective of small data to collect, analyze and mine user scientific research background data as well as and electronic resource behavior data,then divides user electronic resource satisfaction tags in multiple dimensions, on this basis, the user satisfaction portrait of the library electronic resource system is constructed. The portrait can be used to predict changes in user satisfaction, formulate personalized service plans, resource classification recommendations, etc, it can also effectively fit depth services of library and efficiently track the strategic goals of various user characteristics.