Consumption　activities　have　a　great　impact　on　the　construction　and　development　of　urban　commercial　space.　Mining　the　hidden　consumption　activity　patterns　in　spatio-temporal　data　can　help　guide　the　optimization　of　business　scenario　construction.　The　scene　recognition　data　integrates　the　spatio-temporal　information　of　individual　consumers　and　the　semantic　information　of　POI,　and　has　the　advantages　of　high　precision　and　easy　collection,　which　can　be　used　for　large-scale　consumption　activity　research.　In　this　paper,　a　method　framework　of　trajectory　attribute　extraction-consumption　purpose　judgment-space-time　behavior　clustering　is　proposed　by　using　scene　recognition　data.　The　consumption　pattern　is　identified　by　consumption　space-time,　and　the　characteristics　of　consumption　pattern　are　studied　by　combining　consumption　semantics.　Firstly,　a　trajectory　with　consumption　activity　semantics　is　established　based　on　POI　scene　recognition　data.　Secondly,　a　method　for　extracting　the　attributes　of　consumption　trajectory　is　proposed,　and　the　main　consumption　goals　of　individual　consumers　are　further　speculated.　Furthermore,　the　SOM　network　is　used　to　spatially　cluster　the　spatial　and　temporal　attributes　of　different　consumer　individual　activities　to　identify　consumption　patterns,　and　the　characteristics　of　different　patterns　are　studied　in　combination　with　semantic　attributes　such　as　consumption　goals.　Finally,　taking　the　central　urban　area　of　Fuzhou　as　an　example,　this　method　framework　is　applied　to　explore　the　consumption　activity　patterns　of　different　types　of　cities　and　analyze　their　characteristics.　©　2024　Copyright　held　by　the　owner/author(s).