Research On Spam Message Interception System Based On Content Classification

The development of mobile communications technology and the popularity ofthe mobile handsets bring substiantial change to our daily life.Furthermore,thefunction of sending SMS becomes an important part of mobile phones because of itslow cost and convenience.But,along with the growing amount of spam messages,theuser may be harassed,intimidated or even fraud,which is a burdern to both the mobileusers and network operators.Therefore,it is necessary to filter out these spammessages to offer a better user experience.The techniques exploited by networkoperators include:black and white list filtering,content and keyword matching,periodthreshold limit.Although these methods do have some merits,there are still someshortcomings with these existing methods,such as wrong judgement,missing filtering,not real time and difficulty to filter messages sent across provinces.In this thesis,in order to solve the problem of wrong judgement and missingfiltering,a new spam message inception system is proposed,theadvantages/disadvantages of inception location and network element approach arecompared.The design of the system architecture and system software arhictecture,aswell as the flowchart of sending SMS and spam messages inception,are elaborated.In this proposed spam message interception system,a new module of contentclassification,which is based on ICTCLAS is included.Statistic (CHI) is a veryeffective way to reduce the dimension,thus it is adopted for feature selection.Sincesupport vector machine are good at dealing with small samples,have highclassification accuracy,they are exploited to classify spam messages and it is provedthat it can achieve good performance.Finally,comparison experiments are carried out to evaluate the performance ofthe proposed spam message inception system.Performance evaluation follow the relevant indicators of spam filtering.Through results analysis,it is shown that it canachieve high accuracy.