Many　large　enterprises　work　with　thousands　of　suppliers　to　provide　raw　materials,　product　components　and　final　products.　Supplier　relationship　management　(SRM)　is　a　business　strategy　to　reduce　logistic　costs　and　improve　business　performance　and　competitiveness.　Effective　categorization　of　suppliers　is　an　important　step　in　supplier　relationship　management.　In　this　paper,　we　present　a　data-driven　method　to　categorize　suppliers　from　the　suppliers＇　business　behaviors　that　are　derived　from　a　large　number　of　business　transactions　between　suppliers　and　the　buyer.　A　supplier　business　behavior　is　described　as　the　set　of　product　items　it　has　provided　in　a　given　time　period,　a　mount　of　each　item　in　each　order,　the　frequencies　of　orders,　as　well　as　other　attributes　such　as　product　quality,　product　arrival　time,　etc.　Categorization　of　suppliers　based　on　business　behaviors　is　a　problem　of　clustering　high　dimensional　data.　We　used　the　k-means　type　subspace　clustering　algorithm　FW-KMeans　to　solve　this　high　dimensional,　sparse　data　clustering　problem.　We　have　applied　this　algorithm　to　a　real　data　set　from　a　food　retail　company　to　categorize　over　1000　suppliers　based　on　11　months　transaction　data.　Our　results　have　produced　better　groupings　of　suppliers　which　can　enhance　the　company＇s　SRM.