To　find　key　characteristic　attributes,　which　have　deep　influence　on　learning　strategies　in　network　learning,　an　attribute　reduction　method　is　proposed　based　on　rough　set　theory.　The　method　applies　the　discernible　function　principle,　constructs　core　set　and　nocore　set　to　find　the　reduction　through　logic　calculation,　and　sorts　the　attributes　in　the　reduction　set　by　significance　factors　to　obtain　the　key　characteristic　attributes.　The　method　is　applied　to　the　network　English　learning　platform　of　Xi＇an　Jiaotong　University　and　the　result　shows　the　dimensions　of　key　personality　characteristics　are　one-quarter　of　the　original　ones.　The　method　can　find　the　knowledge　of　key　attributes　and　reveal　the　relation　between　personality　characteristic　and　learning　strategy.　Based　on　the　results,　the　original　data　volume　can　be　decreased　to　about　50%.