本文提供了一个从己知连接线交通量等数据直接推导起终点矩阵的方法。其理论涉及到伪逆矩阵,熵极大及最优化方法。同时涉及到在多重解条件下如何确立采样策略,並对起终点矩阵性质有所探讨。具有42个区域,101个节点及861个起终关对的美国宾夕法尼亚州约克市是迄今最大的一个试验网络模型。该市的试验结果证明,与其它类似方法相比,本文论述的方法具有较合理的收敛速度。 This article provides a way to deduce the terminal matrix directly from the data such as the amount of cable traffic. Its theory involves the pseudo-inverse matrix, entropy maximization and optimization methods. At the same time it involves how to establish the sampling strategy under the conditions of multiple solutions and discusses the nature of the terminal matrix. The 42-area, 101-node, and 861-pair Yorktown, PA is by far the largest experimental network model in the world. The city’s experimental results show that compared with other similar methods, the method discussed in this paper has a more reasonable rate of convergence.