The　platform　of　location　based　social　networks　provides　histories　of　users＇　check-in　actions　for　geographical　locations,　namely　Point-Of-Interest　(POI).　Based　on　users＇　location　histories,　we　can　recommend　POIs　in　which　users　may　be　interested.　The　categorical　information　of　POIs　plays　an　important　role　in　POI　recommendations　because　these　information　could　be　used　to　represent　users＇　preference　and　the　properties　of　POIs.　In　this　paper,　we　present　a　category-based　POI　recommender　system　by　drawing　and　extending　results　from　information　retrieval,　especially　the　vector　space　model　and　the　probabilistic　model.　The　location　and　social　information　are　also　fused　in　the　recommender　system.　Experimental　results　show　that　effectiveness　of　recommendations　can　be　improved　by　using　categorical　information.　©,　2015,　Binary　Information　Press.　All　right　reserved.