To　reduce　the　influence　of　implicit　feedback,　data　sparsity　and　content　diversification　on　the　recommendation　algorithm　of　interest　points　and　improve　the　accuracy　of　recommendation,　a　point　of　interest　(POI)　recommendation　algorithm　based　on　sequence　mining　was　proposed　in　this　paper.　First,　in　the　data　preprocessing　stage,　the　negative　sampling　method　was　used　to　generate　data　that　does　not　exist　in　the　data　set　as　negative　samples.　Then,　the　matrix　decomposition　method　was　used　to　learn　the　implicit　feature　vectors　of　users　and　locations,　and　arrange　candidate　recommendation　points　according　to　the　relationship　between　sites.Experiments　on　two　POI　access　sequences　were　implemented　on　the　open　dataset　FourSquare　and　Gowalla.　Results　show　that　the　accuracy　of　the　algorithm　is　much　higher　than　that　of　the　traditional　method.　©　2019,　Editorial　Department　of　Journal　of　Beijing　University　of　Technology.　All　right　reserved.