The　location-based　social　network　(LBSN)　is　expanding　day　by　day,　and　people　are　willing　to　share　their　location　information.　Aiming　at　the　problem　that　the　user-point-of-interest　matrix　is　sparse　and　it　is　difficult　to　obtain　the　user＇s　preference　for　candidate　locations　from　implicit　feedback,　which　affects　the　accuracy　of　point-of-interest　recommendation,　this　paper　proposes　a　matrix　factorization　point-of-interest　recommendation　algorithm　(GMF)　that　integrates　geographic　information.　First,　this　paper　reconstructs　the　user　access　location　preference　matrix　by　using　the　characteristics　of　geographic　information　conforming　to　kernel　density　estimation　to　alleviate　the　problem　of　matrix　data　sparsity;　then,　in　order　to　enhance　the　effectiveness　of　the　matrix　decomposition　model,　we　improve　the　objective　function　of　matrix　factorization　in　the　form　of　an　implicit　feedback　term;　Finally,　it　is　verified　on　the　two　real　datasets,　and　the　results　show　that　the　performance　of　the　algorithm　is　better　than　other　POI　recommendation　algorithms.　©　2023　IEEE.