k-anonymity　is　an　effective　method　of　privacy　preserving.　However,　some　traditional　k-anonymity　models　do　not　capture　diversity　and　dispersibility　of　sensitive　values　in　each　equivalence　class,　which　makes　the　privacy　disclosure　of　anonymity　table　occur　easily.　In　this　paper,　an　advanced　(k,　g)-anonymity　model　for　numerical　data　is　proposed　and　a　(k,　g)-MDAV　algorithm　is　designed　to　achieve　(k,　g)-algorithm.　Experimental　results　show　that　the　algorithm　can　lower　the　risk　of　privacy　disclosure　while　maintaining　the　data　availability.