摘要
研究了改进的基于SVM-EM算法融合的朴素贝叶斯文本分类算法以及在垃圾邮件过滤中的应用。针对朴素贝叶斯算法无法处理基于特征组合产生的变化结果,以及过分依赖于样本空间的分布和内在不稳定性的缺陷,造成了算法时间复杂度的增加。为了解决上述问题,提出了一种改进的基于SVM-EM算法的朴素贝叶斯算法,提出的方法充分结合了朴素贝叶斯算法简单高效、EM算法对缺失属性的填补、支持向量机三种算法的优点,首先利用非线性变换和结构风险最小化原则将流量分类转换为二次寻优问题,然后要求EM算法对朴素贝叶斯算法要求条件独立性假设进行填补,最后利用朴素贝叶斯算法过滤邮件,提高分类准确性和稳定性。仿真实验结果表明,与传统的邮件过滤算法相比,该方法能够快速得到最优分类特征子集,大大提高了垃圾邮件过滤的准确率和稳定性。
This paper discussed improvement of naive Bayesian text classification algorithms based on the SVM-EM algorithms and applications in spam filtering. Naive Bayes algorithm cannot handle the results based on the feature-based combination changes feature-based, and dependent on the distribution of sample space and the inherent instability of the defect, causing the algorithm complexity increases. To solve the above problems, this paper proposed an improved algorithm based on SVM-EM naive Bayes algorithm,which was combined with naive Bayes algorithm＇ s simple and efficient, the advantages of filling the missing property of EM, the advantages of support vector machines （SVM） algorithms, first made nonlinear transformation and structural risk minimization flow into the second classification optimization problem, and then asked the EM algorithm to fill the requirements of the conditional independence assumptions for Bayesian algorithm. Finally, using Bayesian algorithms to improve the mail filtering classification accuracy and stability. Simulation results show that the proposed method can quickly obtain the optimal feature subset classification, greatly improve the spare filtering accuracy and stability compared to traditional methods of mail filtering algorithm.
出处
《计算机应用研究》
CSCD
北大核心
2012年第3期1091-1094,共4页
Application Research of Computers
基金
甘肃省教育科学研究"十一五"规划课题(GS[2010]GX046)
