贝叶斯网络分类器近似学习算法

贝叶斯网络在很多领域应用广泛，作为分类器更是一种有效的常用分类方法，它有着很高复杂度，这使得贝叶斯网络分类器在应用中受到诸多限制。通过对贝叶斯网络分类器算法的近似处理，可以有效减少计算量，并且得到令人满意的分类准确率。通过分析一种将判别式算法变为产生式算法的近似方法，介绍了这种算法的近似过程，并将其应用在了贝叶斯网分类算法中。接着对该算法进行分析，利用该算法的稳定性特点，提出Bagging-aCLL 集成分类算法，它进一步提高了该近似算法的分类精度。最后通过实验确定了该算法在分类准确率上确有不错的表现。

Bayesian Networks Classifier Using Approximation

Bayesian networks are widely used in many fields. As a classifier, it is an effective classification method. Bayesian network classifier is one of the most challenging problems, which makes the Bayesian network classifier subject to many limitations in the application. Through the pairs of Bayesian network classifier algorithms' approximate treatment, it can effectively reduce the amount of calculation, and get satisfactory classification accuracy. This paper analyzes a way to change discriminative score function to generative score function by approximation method. This way is applied in Bayesian network classification algorithm. Finally, this paper uses the stability of new algorithm, proposes a new classifier through integration called Bagging-aCLL. It uses ensemble to improve the accuracy rate of the algorithm. The experiment test shows the classification accuracy rate of the algorithm have a good performance.