中文摘要：

以银行各项资产组合收益最大化为目标函数，以VaR风险收益率为约束，以法律、法规和经营管理约束为条件，运用多期资产分配的逆向递推原理和非线性规划方法，建立了多期银行资产组合动态优化模型．一是通过逆向递推，使本期资产的最优配给建立在下一期资产的最优配给的基础上，解决了现有研究只是在单期里求解而忽略了各期之间的联系与影响的问题．二是考虑到本期的贷款收益率期望值将受到上一期贷款信用等级迁移的影响，利用不同等级贷款收益率期望值和一年期迁移概率矩阵，计算出各类企业各年份的相应的贷款收益率期望值及标准差，更加客观地反映了贷款的真实收益和风险，解决了现有研究只是简单求解各笔贷款收益率期望值或将其设为常数的问题．三是用贷款组合的VaR控制多期贷款的组合风险，解决了现有多期研究中对银行风险承受能力和资本监管的客观要求考虑不足的问题．

英文摘要：

Considering the constrain on VaR, laws, regulations and operation, using portfolio profits maximum of bank as objective function, applying the Backward Induction idea of multi - period assets allocation and Linear Programming Method, the dynamic optimization model of bank asset portfolio is set up. There are some characteristics and innovations in this paper. The characteristics lie on four aspects. Firstly, by using Backward Induction Method, the portfolio of this period is set up based on the portfolio of the next period. And then such problems as the neglect or lack consideration of the interaction of each period can be solved. Secondly, it is taken into account that the earning rate of this period will be affected by loan credit migration of the former period. Then in this study the earning rates of different levels and the probability of risk migration for one year are used to obtain the earning rates of every corporation each year and mean square deviations. It can objectively reflect the actual portfolio and risk. As a result the neglect of the earning rates is avoided. Finally, the portfolio risk of multi-period loan is controlled by the introduction of VaR constrain. Then the lack consideration of the bank＇ s risk tolerance ability and the demand of capital intendance in present multi-period study are avoided.