基于影响因素可靠度的改进OD反推算法

通过4组测试方案，采用敏感性分析、数理统计分析方法分析了先验矩阵总量、分布结构以及观测OD对数量对OD反推结果的影响程度；提出了考虑先验矩阵和观测路段流量可靠度的改进极大熵算法。研究结果表明：先验矩阵总量基本不影响OD反推结果；先验矩阵分布结构决定着反推结果的准确性；OD反推要求观测路段包含所有OD对，且OD对数量越多，反推结果越准确，基本不会出现OD信息冗余；在反推过程中考虑先验矩阵和观测路段流量可靠度，有助于提高反推精度。

Improved OD Estimation Algorithm Based on Influence Factors Reliability

By sensitivity analysis and statistics methods as well as four test proiects, the contribution rates of total volume, distribution structure of prior matrix and influence of observed OD pair numbers on OD estimation result were analyzed. Then, an improved maximum entropy algorithm was put forward in which reliability of prior matrix structure and observed links volume were considered. Research results show that the total volume of prior matrix has little effect on OD estimation result, and the distribution structure of prior matrix plays a decisive role for accuracy of estimation result; OD estimation requires that the observed links must contain all OD pairs, and the estimation result is more accurate when there are more OD pairs; in the process of estimation, considering the reliability of prior matrix and observed link volume can contribute to improve accuracy of estimation result.