Text　categorization　is　one　important　task　of　text　mining,　for　automated　classification　of　large　numbers　of　documents.　Many　useful　supervised　learning　methods　have　been　introduced　to　the　field　of　text　classification.　Among　these　useful　methods,　K-Nearest　Neighbor　(KNN)　algorithm　is　a　widely　used　method　and　one　of　the　best　text　classifiers　for　its　simplicity　and　efficiency.　For　text　categorization,　one　document　is　often　represented　as　a　vector　composed　of　a　series　of　selected　words　called　as　feature　items　and　this　method　is　called　the　vector　space　model.　KNN　is　one　of　the　algorithms　based　on　the　vector　space　model.　However,　traditional　KNN　algorithm　holds　that　the　weight　of　each　feature　item　in　various　categories　is　identical.　Obviously,　this　is　not　reasonable.　For　each　feature　item　may　have　different　importance　and　distribution　in　different　categories.　Considering　this　disadvantage　of　traditional　KNN　algorithm,　we　put　forward　a　refined　weighted　KNN　algorithm　based　on　the　idea　of　variance.　Experimental　results　show　that　the　refined　weighted　KNN　makes　a　significant　improvement　on　the　performance　of　traditional　KNN　classifier.　©　2010　IEEE.