With　the　rapid　development　of　Internet　technology,　indoor　positioning　technology　has　received　extensive　attention　in　recent　years.　According　to　the　growing　demand　for　indoor　positioning,　many　positioning　technologies　have　been　developed.　Among　them,　the　location　method　based　on　Bluetooth　signal　does　not　need　additional　deployment　equipment　and　is　low　cost,　which　has　become　the　focus　of　indoor　location　research.　Bluetooth　signals　are　vulnerable　to　multipath　effects,　and　the　relationship　between　location　information　and　Bluetooth　signal　strength　is　bound　to　be　affected.　The　robustness　and　accuracy　of　most　traditional　machine　learning　localization　algorithms　based　on　Bluetooth　can　not　meet　the　requirements.　Therefore,　in　this　paper,　the　offline　database　is　spatially　divided　based　on　the　artificial　colony　K-means　algorithm,　and　the　online　location　model　is　built　based　on　the　WKNN　location　algorithm　based　on　local　outlier　factor,　and　its　nonlinear　mapping　relationship　with　physical　location　is　established.　The　proposed　method　is　verified　by　collecting　Bluetooth　signal　data　in　a　real　experimental　scene.　The　experimental　results　show　that　the　proposed　method　effectively　reduces　the　interference　of　noise　and　achieves　the　best　accuracy.　©　2023　IEEE.