針對中文改善垃圾信件過濾準確度之研究

Abstract

垃圾信件在網際網路中已經成為了極大的威脅，不僅浪費網路資源，也浪費使用者的時間。本篇論文分析製造垃圾信的各種方法與阻擋垃圾信件的各種過濾機制，此類過濾方法最有名的是採用機率與統計的分析。由於大部分的過濾垃圾信件系統只是處理英文信件，於是，論文將針對中文垃圾信，採用Bayesian classifier的方法過濾中文垃圾信件。從實驗結果可以得知這個方法可以提升過濾的準確度。

Spam is now become a serious threat to the Internet. This thesis examines the problems and impact caused by spam. The thesis also examined methods and techniques to generate spam, and examined methods to classify spam. There are many filtering algorithms introduced to fight spam. One of the most popular filtering algorithms is statistical based filtering, which is based on Bayesian classification theorem. However, these algorithms focus on the English e-mails. This thesis presents a Chinese e-mail classifier, which is based on Bayesian classifier to classify Chinese spam e-mails. The experiment results show that the proposed Chinese e-mail classifier could perform a high accuracy.