Research On Spam Filtering Of Particle Swarm Optimized SVM

With the popularization of Internet e-mail as a convenient fast and cheap way of communication has been a great popularity. E-mail has brought great convenience to people’s work and life, it also makes the whole networks are overwhelmed with the overflow of spam, occupy the network bandwidth, invade and occupy recipients email storage, consume recipients much time and it is common occurrence of tens of thousands of spam assault the target website to make the target website network blocking and ends up paralyzing. Spam reduces the efficiency of the entire network, also caused serious harm to the society. How to prevent and filter spam has aroused people’s great attention, which is also one of the hot research in the field of network security.This paper also learn research and disscuss for spam filtering, using particle swarm optimized support vector machine (PSO-SVM) for spam filtering, and with the other algorithms such as Naive Bayes, support vector machine (SVM) and genetic algorithm optimized support vector machine (GA-SVM) are theoretically horizontal comparison proves that PSO-SVM is more suitable for Chinese spam filtering than other commonly used in spam filtering algorithm, the experimental results also proved this point.The content of this paper can be divided into four parts. First of all, introduced the research background significance research status at home and abroad of spam filtering and the main research work of this paper, and the content arrangement of each chapter.Secondly, briefly introduced the basic knowledge of e-mail, mainly including the working principle of e-mail, the main protocols used in the transmission of email, the definition of spam and it’s classification and the pretreatment process of Chinese e-mail.The third part gives three kinds of common spam filtering technologies, respectively includes based on IP layer filtering, based on SMTP protocol filtering and content-based filtering. This article mainly conducts the filtering research based on content, and gives four kinds of frequently-used content-based spam filtering algorithms. This paper proposes using particle swarm optimized support vector machine for Chinese spam filtering, and we can get the conclusion by theoretical analysis PSO-SVM can recongnize the spam more quickly and accurately than NB SVM and GA-SVM these commonly used spam filtering algorithms.The last part is the experiment of this paper, gives the experiment result and it’s performance analysis. By comparing the results of experiment get the conclusion:in the same experimental condition SVM has better filtering performance index (including recall rate, accuracy rate, precision rate and F value) than NB in the Chinese spam filtering. The filtering performance of parameters optimization of SVM has improved than that without parameters optimization of SVM. The filtering performance of PSO-SVM is much better than NB, SVM and GA-SVM filtering algorithm, and the evolutionary algebra of particle swarm optimized support vector machine parameters are shorter than genetic algorithm optimized SVM parameters, and the optimization result is much better.