Clustering　is　used　to　identify　the　distribution　pattern　of　the　data　set　based　on　the　similarity　of　data,　but　the　relationship　between　data　is　ignored　in　the　most　existing　clustering　processes.　This　paper　reveals　the　production　relationship　between　inputs　and　outputs　from　the　evaluation　perspective　of　decision-making　units　(DMUs),　and　innovatively　introduces　data　envelopment　analysis　cross-efficiency　approach　to　construct　a　new　clustering　approach.　This　new　approach　not　only　can　cluster　DMUs　based　on　the　production　relationship　between　data,　but　also　can　reflect　the　preference　of　decision　maker.　The　clustering　results　are　relatively　stable　and　unique,　and　they　are　meaningful　for　analyzing　DMUs　in　production　activities.　In　addition,　the　new　cross-evaluation　strategy　based　on　the　nearest　neighbor　is　proposed　to　further　optimize　the　clustering　process　by　considering　data　characteristics,　and　then　more　reasonable　and　objectively　clustering　results　can　be　obtained.　Finally,　two　examples　are　provided　to　illustrate　the　effectiveness　and　practicability　of　the　new　clustering　approach.　©　Operational　Research　Society　2020.