摘要
针对网络文本信息的安全性判别问题,采取改进的邻近分类算法挖掘文本.该改进邻近分类方法在传统方法定义分类特征的同时,起用共线性判别矩阵,对具有共线属性的特征合并处理.这种改进策略,不仅可以增加分类特征的准确性,也可以加快文本信息的分类进程.对Spambase语料库开展实验研究,从精度、召回率、联判度、误差4个维度对分类效果进行评价.结果显示:改进的邻近分类方法具有明显的优势,可以更加准确地区分安全文本和危险文本.
In view of the security problem of network text information, we adopt an improved neighbor classification al- gorithm to carry out text mining. In improved nearest neighbor method, definition and classification are carried out by tra- ditional method, and characteristics are merged by reinstating co-linear discriminant matrix of collinear attribute features. This improved strategy not only increase the accuracy of classification features, but also speed up the classification process of text information. An experimental study is carried out on the Spambase corpus, and the classification results are evalu- ated from 4 dimensions. Namely accuracy, recall rate, the degree of error, and the error. Results show that the improved method has obvious advantages, and that is more accurate in the area of security text and dangerous text.
出处
《华侨大学学报（自然科学版）》
CAS
北大核心
2016年第1期67-70,共4页
Journal of Huaqiao University(Natural Science)
基金
四川省自然科学基金重点资助项目(15ZA0349)
关键词
文本信息
文本挖掘
文本分类
邻近分类
text information
text mining
text classification
neighbor classification
