Research On Content-based Short Message Intelligent Analysis System

In recent years, spam short massages question was serious day by day, not only for telecommunication operation business, also gave the entire society to have the very tremendous negative influence. Each telecommunication operation business uses each kind of spam short massages filtration system, governs spam short massages being in flood. At present mainly methods of using trash short note determination includes: Judgment user's sampling frequency, judgment short massages content whether contains the key words and the basis called subscriber number characteristic determination. Although these methods have certain effect to spam short massages government, but also has the obvious flaw and the insufficiency.This dissertation in view of the existing short messages filtration technology insufficiency, studies the effective solution or the improvement method, in the original filtration technology foundation, proposed more effective judges spam short massages based on the content of short massages semantics intelligence analysis method, This main contents and contributions are as follows:1. Discussed spam short massages filtration question research present situation, summarizes trash short note judgment mechanism and advantage and shortcoming which present operation business mainly uses.2. Designed one content-based short massage intelligence analysis system, has formulated the system overall construction, various modules main function and the system processing flow.3. Realized the frequency threshold analysis, key words processing and the content intelligence analyzes three main function modules, proposes one kind of specific time recollection algorithm, the statistical user transmission spam short messages frequency.4. Proposed one kind the classified algorithm which unifies the na?ve bayes algorithm and support vector machines. The Na?ve Bayes algorithm has the speed quickly, the efficiency high characteristic, available in online classification. But the support vector machines precision is high, when processes the large-scale short message text, the convergence rate is slow, available in off-line classification, and by through the feedback, renews the short messages characteristic sample database.Proved through the experiment that compares the traditional methods, use improvement algorithm design realization short message intelligence analysis system, success enhancement short message classification intelligence, reliability, timeliness and rate of accuracy. This dissertation proposed content-based short message intelligence analysis system is the improvement and the consummation in the intellectualized aspect to the current short message filtration system, has the extremely broad project application prospect.