Clustering　is　an　unsupervised　method　for　data　analysis,　and　cluster　validity　is　largely　dependent　on　the　estimation　of　the　number　of　clusters.　In　this　paper,　we　have　studied　the　feature　of　high　dimensional　data,　give　a　method　to　calculate　the　compactness　of　intra-cluster　and　isolation　of　inter-cluster　using　geodesic　distance,　and　propose　a　manifold　approach　for　cluster　validation　of　high　dimensional　data.　The　experimental　result　shows　that　the　new　validation　approach　works　better　than　original　one　on　high　dimensional　dataset.　1553-9105/　Copyright　©　2009　Binary　Information　Press.