CAJ | 학술논문

传统的贝叶斯垃圾邮件过滤系统虽然具有较高的分类准确性，但是在处理邮件时存在效率低、消耗资源量大的问题。本文针对贝叶斯垃圾邮件过滤算法进行了在 Hadoop MapReduce 下的研究，并对判定类别的阈值进行了优化，实验表明，本文提出的算法降低了正常邮件的误判率，提高了垃圾邮件判定的准确率和 F 值，同时提高了垃圾邮件过滤的效率。
전통적패협사랄급유건과려계통수연구유교고적분류준학성，단시재처리유건시존재효솔저、소모자원량대적문제。본문침대패협사랄급유건과려산법진행료재 Hadoop MapReduce 하적연구，병대판정유별적역치진행료우화，실험표명，본문제출적산법강저료정상유건적오판솔，제고료랄급유건판정적준학솔화 F 치，동시제고료랄급유건과려적효솔。
Although the traditional Bayesian spam filtering system has a high classification accuracy , but the problems of low efficiency, needing to consume a lot of resources are exiting in the processing of mail. In this paper, the research on Hadoop MapReduce for Bayesian spam filtering algorithm is carried out, and the threshold of determine category is optimized. Experiments show that the algorithm proposed in this paper reduces the error rate of legitimate emails , improves spam email accuracy and F value. At the same time, it improves the efficiency of spam filtering.