Research On Hybrid Swarm Intelligence-Based Spam Filtering

As e-mail is widely used, more and more spam messages appear. And the biggest chiefculprit is the manufacture and disseminator of harmful information. They use email whichis a popular communication tool between people to make and to disseminate various spammail widely, these spam more or less contain trash information like the pornography, thecheating, reactionary and malicious advertisement. This bad practice not only causingserious congestion to the network, but brought many unstable factors to the society, itsimpact is extremely bad.The poor security defect of email has plagued many countries and scholars, in order toremedy this defect and give network a fine green environment, Many countries have issueda number of legal documents, through legal means to curb the generation and disseminationof spam. However saves a life must seek for fundamental way, blindly through the law tosuppress its effect is not obvious, we should work hard in network source, In such an urgentdemand,generated a variety of different types of spam filtering technology,among them,The email filtering which based on the content of the email has been used widely andworthy of study, This article has mainly studied email filtering based on the email content,to delete irrelevant and redundant characterized by the feature selection, and to reduce thesize of the feature subset and improve the classification accuracy, The reason to study thisfiltering technology is that technology continues to progress, more and more peopleunderstand the mysteries of the life. As an important subject of the mysteries of life. Lifescience continually favored by scholars, the results of various researches are numerous andpresent, many are beginning to get rid of logical calculation rules in the past, began toexplore new methods of calculation. In such an environment, biological masscharacteristics behavior caused a great interest to the researchers, so they set mathematicalmodel of these features and simulate it on the computer, so as a new discipline, swarmintelligence (Swarm Intelligence, SI) came into being.In the last century90’s, Italy of scholars M Dorigod proposes a new of intelligentalgorithm-Ant group optimization algorithm (Ant Colony Optimization, ACO), just likeNewton inspired by apple fell to found gravitation, after carefully observation on Ant, MDorigod found that some of ants in the ant colony take the shortest path during foraging, soAnt group optimization algorithm above was proposed. Now that the algorithm has beenapplied to data classification, data clustering, pattern recognition, biological modeling andso on. Almost at the same time J Kennedy and R C Eberhart also introduced a new type ofparticle swarm optimization algorithm (Particle Swarm Optimization, PSO). thealgorithm is also based on the observation of the behavioral characteristics of the cluster animal like ants, birds, fish. Because of the advantage of simple concept, few parameter,and easy to implement, this algorithm raise extensive attention by researchers at home andabroad and soon applied to many other fields.This article proposed a mixed group of intelligent spam filtering methods (HybridSwarm Intelligence Spam Filtering Method, HFM) combine these two kinds of intelligencealgorithm, in view of spam mail data. For feature subset search problem, proposes a fuzzymultiple-population Particle Swarm (Fuzzy Multi-swarm Particle Swarm Optimization,FMP) and an improved ant colonies optimization (Improved Ant Colony Optimization,IACO). HFM is composed of three phases: at the first stage adopt the redundantinformation gain and FMP search features; the second stage used FMP find core features;at the last stage complete features selection with IACO. Test by PU1, Ling-Spam, SpamAssassin Data Set. comparison study shows: HFM can search to a smaller subset of features,can obtain higher classification accuracy, thus enhances the performance of spam detectionfilters, have the potential for applications and continue to study.