关键词: 朴素贝叶斯分类, 参数嵌入, 文本分类, 后验概率, 分类可视化

Abstract: How to visualize the text classifier result is one of the focus field in pattern recognition.On the assumption that each class can be represented by a Gaussian distribution in the embedding space，through Naive Bayes classification algorithms posterior probability for data over classes was got，Parametric Embedding（PE） algorithm was applied into the visualization of classification result in low-dimensional.PE algorithm tries to preserve the structure in an embedding space by minimizing a sum of Kullback-Leibler divergences in high-dimensional space.Data that are located at the center of cluster are typical data for the class，and data that are located between clusters have multiple topics，different data are located in the cluster of different classes.The outstanding advantage is that computing complexity is just the type of data and the corresponding number of the product，is well suited to large volume of data，fewer types of classified data visualization.Experimental result on 20 Newsgroups data sets and MiniNewsgroups data sets proves the effectiveness of the method.

Key words: Naive Bayes classifier, parametric embedding, text classification, posterior probability, classification visualization