【摘要】 针对垃圾邮件文本数据高维、稀疏及词条相关等特点,提出Elastic Net-Decision Tree(EN-DT)两步分类算法。第一步,利用Elastic Net提取邮件文本特征变量,将高维文本数据降至低维。第二步,将所提取的低维特征变量输入到Decision Tree中进行邮件分类。根据分类评价指标对分类结果进行评价。利用Mark Hopkins等人收集的Spam邮件文本数据进行仿真,实验结果表明相比于PLS、PCA和Lasso等算法EN-DT分类性能更佳。更多还原

【Abstract】 A classification algorithm based on Elastic Net-Decision Tree( EN-DT) is proposed,which is suitable for the email text data with characteristics such as high dimension,sparseness and correlation. Firstly,the email text characteristic variables are extracted to make the high-dimensional text data to the low ones through Elastic Net algorithm. Secondly,the low variables are used as the input of the decision tree in order to classify the emails. And the classification result is evaluated according to the classification evaluation index. Finally,simulation studies are implemented using the Spam dataset collected by Mark Hopkins. The results show that the performance of the EN-DT algorithm is more better than the PLS、PCA and Lasso.更多还原