利用类概率估计的加权平均树增强朴素贝叶斯网络结构

Improving weighted averaged tree naive Bayes based on class probability estimation

树增强朴素贝叶斯模型通过放松条件属性独立来改进贝叶斯模型,结构学习效率较高且简单。然而在一些实际实验测试中,树增强朴素贝叶斯分类模型的分类精确性和失误率的效果却不好。因此,设计了加权平均的树增强朴素贝叶斯分类算法来改进分类的效果,并且利用对数条件似然函数来测试分类估计的效果,给出加权平均的树增强朴素贝叶斯分类模型在训练阶段和测试阶段的算法,最后利用Weka平台公布的大量的UCI数据集通过十字交叉验证法来进行实验,结果表明加权平均树增强朴素贝叶斯分类模型明显优于最优朴素贝叶斯分类模型和树增强的朴素贝叶斯分类模型。

Tree augmented naive Bayes（ TAN） improves naive Bayes（ NB） by weakening its conditional attribute independence assumption,while maintaining efficiency and simplicity. In many real-world applications,however,classification accuracy or error rate of TAN was not enough. Thus,this paper investigated weighted averaged tree naive Bayes（ ATAN） algorithm to improve its class probability estimation performance. And it estimated performance of ATAN in terms of log conditional likelihood（ LCL）. Meanwhile it applied the algorithm of the training and testing to estimate the performance. The experiments were done on a large number of UCI datasets published on the main Web site of Weka platform by using the methods of cross validation.The results show that ATAN significantly outperforms TAN and ONB all the other algorithms used to compare in terms of LCL.