Research On Technology Of Spam Filtering Based On Markov Logic Network

E-mail has been widely used all around the world since 1980s. It is fast, convenient and inexpensive, which makes it popular among internet users. With the massive popularity of the Internet, E-mail has become a common means of communication. No matter personal communications or business communications, E-mail can always play an important role. E-mail brings us a lot of conveniences, but meanwhile it brings us trouble. Because it is low-cost, many advertisers send spam to internet users and bring us many inconveniences. Many fraudulent spam even cause huge losses. How to filter spam has become a public topic on Internet. Consider the development of spam filtering technology, many filtering methods has been used chronologically, they are methods based on key character, black list, rules and contents. These methods have their own advantages. Within a certain range of applications, each method has perfect performance. However, the senders of spam make many new kinds of spam to avoid these filtering systems. The existing filtering methods can't solve it, and there are spam undetected or normal e-mails false detection.After learning the latest scientific and technological achievements on spam filtering adaptability, we analyze how to improve the spam filtering adaptability in this paper, and the research results are shown below:1. we systematically summarize the main technologies on spam filtering and make comparison on the technologies. After analyzing the advantages and disadvantages of each method, we point out that the adaptability based on Markov Logic Network is worth deeply study, it is an important method to improve the adaptability.2. we deeply analyze the relevant theories on Markov Logic Network and propose a method on how to use Markov Logic Network relevant theories to improve spam filtering adaptability. Furthermore, we make a rigorous proof and validation on adaptability boundary. 3. We realize a spam filter based on the theories above and verify with actual collected data. After compared with filters using other methods, we prove the correctness of our method.