Recommender system (RS) suggests useful items to users. Most of existing techniques in RS focus on using a rating matrix of users and items without considering recommendation context. In this paper, we take advantage of multidimensional database to present context information in context-aware RS. Therefore, we exploit the ability of OLAP aggregate operations to estimate user ratings. We propose an approach to translate the concepts of content-based, collaborative filtering and context recommendation into OLAP aggregate operations and integrate them to rating estimation function. Furthermore, through OLAP aggregate operations, our rating estimation function tends to solve the cold-start problem in RS and the data sparsity problem in context-aware RS. We develop a context-aware tour RS with our approach. In this system, we survey related researches to identify context information and user, item information being suitable to tour RS. We evaluate the system by accuracy and performance.