结构化EM算法在小样本贝叶斯网络学习中的应用

Application of structural EM algorithm to learning Bayesian networks for small sample

在贝叶斯网络学习中,合理数据集的存在可以大大降低贝叶斯网络学习对知识工程的过多依赖.但当数据集中样本数量不够大时,可能没有足够多的样本甚至不存在样本来代表变量间的某些条件独立关系,从而无法学习贝叶斯网络.将数据集修正与结构化-期望最大化算法相结合,得到一种有效的小样本上贝叶斯网络学习的方法,实验结果表明,该方法能够有效地进行小样本上贝叶斯网络学习.

Existing data sets of cases can significantly reduce the knowledge engineering effort required to learning Bayesian networks.When a data set is small,many conditioning cases are represented by too few or no data records and they do not offer sufficient basis for learning Bayesian networks.It is proposed a method that combines data revising and the Bayesian Structural EM algorithm.Experimental results show that this method is effective in learning Bayesian networks from small data set.