指出了翻转决策树法求解完全信息价值时可能存在问题的根本原因，提出完全信息价值在影响图中的本质表现为信息弧，并对Howard正则型影响图定义进行修正，进而系统分析了两种不同情形下基于影响图求解完全信息价值的思路。最后举例说明。

The problem of calculating the value of perfect information based on the decision tree is analyzed. Then, the article presents that the value of perfect information could be described by the information arc in the influence diagram. Also, the definition of the Howard canonical form is modified. Last, how to calculate the value of perfect information based on the influence diagram is analyzed systematically.