In event-based social networks (EBSN), groups that aggregate users with similar interests for sharing events play important roles in community development. Understanding why people join a group and how groups are formed is particularly an interesting issue in social science. In this paper, we study predicting users' preferences on social groups by considering content information in EBSN, i.e., geographic-social event-based recommendation. Specifically, we consider two types of content information, i.e., the tags and geographical event locations about users/events. We propose the SEGELER (pair-wiSE Geo-social Event-based LatEnt factoR) to model the users behavior considering the information. Experiments on a real-world EBSN social network validate the effectiveness of our proposed approach for both normal users and cold start users.