Research On Multi-level Spam Short Messages Filtering System Based On Text Classification

In recent years, mobile phone short messages favored by the majority of users, spamshort messages generated by this business seriously troubled people’s lives, interfered withthe normal social order. Therefore, the telecommunication operators have introduced spamshort messages filtering system to fight spam short messages serious flooding. Now,commonly used filter methods are mainly with black and white list filtering mechanism,based on short messages’ length and flow limit filtering mechanism, based on key wordfiltering mechanism, as well as based on text classification filtering mechanism. To someextent, these mechanisms have curbed the spam short messages flooding, but there are stillobvious flaws and shortcomings.This article is inspired by the biological immune system, referenced biologicalimmune principle, presented a multi-level spam short message filtering system based ontext classification. The system is divided into several modules, layer by layer filter, linkedtogether, nowhere to hide the spam short messages. The work done in this article, as wellas innovation points as follows:(1) This article analyzed the status of spam short messages, including causes, damagesand the main features. To domestic and international, the existing spam short messagesfiltering mechanism are summarized. It discussed the advantages and disadvantages ofvarious mechanisms.(2) This article introduced the key technology of spam short messages filtering,including text pretreatment, Chinese word segmentation and the feature selection methods.It elaborated on the principle of biological immune system and the development ofartificial immune algorithm. The efficiency of the algorithm is improved. The antibodiesand antigen expression methods are simplified. These make the calculation of affinity moresimple, save the memory, improve the matching speed.(3) This article designed a multi-level spam short messages filtering system based ontext classification. The system includes the black and white list module, message lengthand flow threshold module, content fast matching module and artificial immune module. Itformulates the main funtion and processes of each module.(4) Artificial immune module via the training, antibodies via the autologous tolerancegeneration detector achieve to filter the short messages. Short messages characteristics ofthe sample library can also be continuously updated through the variation mechanism toensure the diversity of the sample database and adaptability. The experiment results show that, compared with the traditional method thatconcerned, multi-layered spam short messages filtering system based on text classificationis intelligence, reliability, accuracy and real-time. It is an intelligent improve and perfect,with very broad application prospect.