一种改进的朴素贝叶斯不平衡数据集分类算法

An improved Naive Bayes classification algorithm for unbalanced data sets

当训练集中各个类别的样本分布不均匀且存在数据稀疏问题时,朴素贝叶斯算法分类不够准确。针对此问题,提出了一种基于数据平滑与加权补集的朴素贝叶斯文本分类算法,该算法引入数据平滑算法计算贝叶斯模型中缺失特征的补偿概率,克服数据稀疏问题;利用当前类别补集的特征来表示当前类别的特征,解决训练集中各个类别的样本分布不均匀时,分类器容易倾向于大类别而忽略小类别的问题。实验结果表明,在样本集分布不均衡时,该算法比传统的朴素贝叶斯分类算法分类效果更好。

When training samples of each class are distributed unevenly and sparsely,the classification efficiency of Naive Bayes is not accurate enough. To solve this problem,a Naive Bayes text classification algorithm based on data smoothing and weighted complementary set was proposed,using data smoothing algorithm to calculate the compensation probability of the missing feature in Naive Bayes model,which can solve the data sparseness problem. Since training samples of each class are distributed unevenly,it uses features of current categories＇ complementary set to represent the features of current categories,which can solve the problem of recognizing the larger category and ignoring the smaller category. The experimental results show that the classification efficiency of the proposed algorithm is better than the traditional Naive Bayes when the training data set is uneven.