In　this　paper,　a　fuzzy　clustering　approach　forTS　fuzzy　model　identification　is　presented.　In　the　proposed　method,　the　modified　mountainx　clustering　algorithm　is　employed　to　determine　the　number　of　clusters.　Secondly,　the　fuzzy　c-regression　model　(FCRM)　algorithm　is　used　to　obtain　an　optimal　fuzzy　partition　matrix.　As　a　result,　the　initial　parameters　can　be　determined　by　the　optimal　fuzzy　partition.　Finally,　gradient　descent　algorithm　is　adopted　to　precisely　adjust　premise　parameters　and　consequent　parameters　simultaneously.　The　simulation　results　reveal　that　the　proposed　algorithm　can　model　an　unknown　system　with　a　small　number　of　fuzzy　rules.　©　Springer-Verlag　Berlin　Heidelberg　2014.