Junk mail classification method based on partial match estimation

Abstract

The invention discloses a rubbish mail filtering method based on partial matching forecast, comprising: 1) transform a novel mail into a character string composed of corresponding characters of 001-127ASC II; 2) extracting ancient rubbish mail training set, a normal mail training set, a rubbish mail forecast set and a normal mail forecast set; 3) training the normal mail training set and the rubbish mail training set into a rubbish mail model and a normal mail model through a partially matching forecast algorithm; 4) performing cross entropy operation on the character string and the rubbish mail model transformed by a novel mail with the normal mail, obtaining two cross entropy values; 5) determining whether the novel mail is the rubbish mail or the normal mail by the model with the smallest cross entropy; 6) after sorting the novel mail, adding the novel mail into the forecast set, performing incremental learning, obtaining a novel model. The invention effectively avoids the condition that the normal mail is used as the rubbish mail by mistake.

Description

technical field [0001] The invention relates to a spam filtering method, in particular to a spam classification method based on partial matching prediction. Background technique [0002] The popularity of e-mail is mainly due to its convenience, speed and low cost; with the popularization of the Internet, e-mail has gradually become one of the convenient means of communication in people's lives. However, in recent years, with the vigorous promotion of electronic informatization in large traditional industries, spam in the field of information systems has inevitably increased exponentially. Spam has the following characteristics: large number, repetitive, coercive, deceptive, unhealthy and fast spreading. So it seriously interferes with people's normal life and poses a serious threat to the information network. Since the types of spam are becoming more and more complex and diverse. Therefore, studying spam classification has become an important research topic in recent yea...

AI Technical Summary

Problems solved by technology

[0008] However, these three methods have problems in terms of speed and efficiency, and none of them are incremental learning methods, and there are doubts in terms of accuracy.

Images