A learning algorithm of compressed candidates based on Bayesia belief network is developed to solve slow running problem of traditional Bayesian belief network constructing algorithm.

• 摘要针对传统算法分类速度较慢的不足，改进传统算法中候选变量的搜索方式，提出用依赖度量函数测量变量之间的依赖程度，得出压缩候选的贝叶斯信念网络构造算法。