【摘要】 为实现朴素贝叶斯分类模型的无监督学习,提出一种基于数据挖掘理论中聚类算法思想的学习方法。该方法首先定义不同类型单维状态分量的差异度量方法和混合型多维向量的联合差异度量方法,通过分析样本数据中向量之间的差异性进行聚类统计,得到研究对象的分类类别,然后对各单维状态分量分别聚类得到特征核值,进而确定不同类别各分量对应单维特征核值空间的概率隶属度。仿真实验结果表明,该方法能有效进行朴素贝叶斯分类学习。更多还原

【Abstract】 To achieve unsupervised learning of naive Bayesian network classifier,a method based on clustering thinking from theory of data mining is proposed. Difference measurement of single-dimension and multi-dimension associated vector of different date type is defined,type set of vector and kernels of its each components are made from clustering with difference measurement of sample data,and then determine the probability of membership grade from every component of vector to the counterpart kernels. Practical example shows that the method can be applied to the learning of naive Bayesian network classifier effectively.更多还原