This　paper　uses　CVaR　as　a　measure　of　risk　to　study　portfolio　investment　of　urban　real　estate　and　establish　an　optimization　model　for　the　portfolio　investment.　Through　this　model,　the　loss　of　a　real　estate　investment　and　the　optimum　portfolio　within　a　single　urban　area　can　be　calculated　at　a　given　confidence　level.　Then　this　model　is　applied　to　calculate　the　risk　and　the　investment　portfolio　based　on　indexes　of　real　estates　of　the　different　districts　and　types　in　the　city　of　Shenzhen　from　June　2006　to　August　2008.　The　empirical　study　shows　that　the　risk　obtained　by　the　optimal　model　based　on　CVaR　is　better　than　the　one　based　on　VaR.　The　main　risk　control　strategy　is　then　to　choose　the　investment　portfolio　that　has　low-risk　CVaR　so　as　to　reduce　project　risk　and　to　improve　the　stability　of　return　on　investment.　©2009　IEEE.