中文摘要：

考虑交易成本和交易量限制，提出投资组合的收益率的隶属函数为梯形的多阶段均值一半绝对偏差可能性投资组合模型，并用自创算法——离散近似迭代法求解。其基本思路为：将连续型状态变量离散化，根据网络图的构造方法将上述模型转化多阶段赋权有向图；运用极大代数求出起点至终点的最长路程，即获得模型的一个可行解；以该可行解为基础，继续迭代直到前后两个可行解非常接近。文章还证明了该方法的线性收敛。最后，文章以一个具体的算例验证了该算法的有效性。

英文摘要：

Considering the transaction costs and the constraints on trade volumes, the paper proposes the multiperiod mean semi-absolute deviation fuzzy portfolio selection model which the fuzzy return rates are denoted as trapezoidal fuzzy numbers and the new algorithm -- the discrete approximate iteration method is proposed to solve the model. The algorithm is following. According to the network method, discredits the state variables and transforms the model into multiperiod weighted digraph; uses max-plus algebra to solve the maximal path that is the admissible solution continues iterating until the two admissible solution is near based on the admissible solution. At last, the paper proves the linear convergence of the algorithm. At last, we prove the algorithm is highly effective by an example.