Correlation　clustering　problem　is　a　classical　clustering　problem　and　has　many　applications　in　protein　interaction　networks,　cross-lingual　link　detection,　communication　networks,　etc.　In　this　paper,　we　discuss　the　capacitated　correlation　clustering　problem　on　labeled　complete　graphs,　in　which　each　edge　is　labeled　+　or　-　to　indicate　two　endpoints　are　＂similar＂　or　＂dissimilar＂,　respectively.　Our　objective　is　to　partition　the　vertex　set　into　several　clusters,　subject　to　an　upper　bound　on　cluster　size,　so　as　to　minimize　the　number　of　disagreements.　Here　the　number　of　disagreements　is　defined　as　the　total　number　of　the　edges　with　positive　labels　between　clusters　and　the　edges　with　negative　labels　within　clusters.　The　main　contribution　of　this　work　is　providing　a　5.37-approximation　algorithm　for　the　capacitated　correlation　clustering　problem,　improving　the　current　best　approximation　ratio　of　6　[21].　In　addition,　we　have　conducted　a　series　of　numerical　experiments,　which　effectively　demonstrate　the　effectiveness　of　our　algorithm.