Fejer Means Application for Determination of Mail Document Spam Level
Korobeynikov Anatoly Grigorevich, Gatchin Yuri Armenakovich · 2014
In this paper we propose algorithms for mail documents spam level determination and training sample development. The problem of spam level determination is based on support vectors method. The modification of the standard method is based on construction means separating hyper planes and on Fejer means. Unlike other algorithms, this algorithm allows to work with non-stationary data, which are used for documents classification. Information technologies generated the continuously artificial intelligence, the application of mathematical increasing flow of diverse information. The main task of statistics, etc. (1, 2). the search systems (search engines) is to provide One of the most popular filters nowadays is the filter qualitative results, i.e. the most important relevant pages. which is based on the Bayesian approach (naive Bayesian It is necessary to solve the problem of classification for classifier), which assumes that different terms of such a provision. Therefore, the theory, the methods and messages are independent of each other. The semantic the algorithms of information classification are rapidly relations between terms should be considered to improve developing scientific trend. the effectiveness of such filters and it requires the use of One of the major problems met by almost every semantic analysis techniques, which significantly Internet user is a spam problem, i.e. the problem of increases a system load and filter work time, with a slight incoming information filtering (classification). increase of filtration efficiency. Currently, a number of filtering technologies - Thus, taking into account the abovementioned facts, services is developed to avoid unwanted information. there is a need to develop new methods and algorithms These technologies can be divided into manual and for information classification to solve spam filtration