Carbon　emissions　drive　climate　change.　Especially　with　the　rapid　development　of　economy,　carbon　emissions　are　increasing　in　recent　years,　and　the　carbon　emission　data　sets　are　more　comprehensive.　How　to　analyze　the　data　is　important.　Furthermore,　to　find　the　main　characteristics　of　carbon　emission,　we　propose　a　new　method　of　segmentation　in　the　time　series　that　adopts　communities　finding　in　complex　network,　graph　convolution　networks　(GCN)　and　visibility　graph　(VG).　Experiments　on　carbon　emission　datasets　show　that　the　detector　has　better　detection　performance　than　existing　graph　connectivity-based　detectors.　In　addition,　we　find　that　combining　the　results　of　GCN　segmentation　can　highlight　economic　geographic　attributes　such　as　resource　endowment,　industrial　structure,　and　market　demand　in　carbon　emission　regions,　thus　complementing　the　existing　applications　of　complex　network　methods　in　the　energy　field　and　providing　insights　for　decision　support　of　carbon　emissions.