摘要
本文提出了短信语义和号码特征相结合的垃圾信息智能识别方法。在分类器的设计上,采用了BP神经网络与支撑矢量机(SVM)的分类集成技术,使得分类识别效果明显。垃圾短信的学习样本识别正确率达99.86%,测试样本识别正确率达到97.4%。由于本文方法提取的特征构成了稀疏矩阵,因此大大缩短了机器学习时间,使得系统具有实时学习和实时提高分类能力的功能。
In this paper, an intellectual recognition method with the features of the number combined with short message＇s semantic is taken out. Considering the design of collecting classifier, both the BP neural network and the support vector machine （for short SVM） are adopted. Therefore the effect of classifying and recognizing is obvious. Correct rate reaches 99.86% as for learning samples and is up to 97.4% as for testing samples, Based on the features forming sparse-matrix, learning time is shorter and the function of real-tlme learning and improving the classifying ability is held.
出处
《电信科学》
北大核心
2008年第8期61-64,共4页
Telecommunications Science
关键词
垃圾短信
BP神经网络
SVM
机器学习
junk short message, BP neural network, SVM, machine learning
