贝叶斯网络分类器,已经广泛应用于各种分类问题.对于固定的贝叶斯网络结构,可以通过生成参数和判别参数2种方法进行学习.生成参数学习效率较高但是分类精度较低,而判别参数学习与之相反.通过对数据集中参数出现频率计算来进行参数学习,并加入一个判别参数来加强实例属性与分类目标值之间的关联性,在此基础上提出了一种简单、快速、有效的判别频率估计(DFE)算法.实验结果表明:在油水层模式识别当中,这种判别频率估计方法相对于其他算法在分类精度上能够提高5%~10%.%Bayesian network classifiers have been widely used for classification problems. Given a fixed Bayesian network structure, there are two different approaches of parameter learning; generative parameter learning and discriminative parameter learning. Generative parameter learning is more efficient but discriminative parameter learning is more effective. We propose a simple efficient and effective discriminative parameter learning method. It is called discriminative frequency estimate (DFE) algorithm which adds a discriminative parameter to strengthen the correlation between properties and classification. It learns parameters by discriminative calculating frequencies of parameters from data set. The experiment shows that the accuracy of discriminative frequency estimation method is 5% -10% higher than the other methods.
展开▼