中文摘要：

分析了目前在垃圾邮件过滤中广泛应用的基于最小风险的朴素贝叶斯模型，提出了一种新的基于直线几何分割的朴素贝叶斯邮件过滤模型LGDNBF，定义了新的风险因子。新的风险因子对决策风险的描述更加精确，同时使得LGDNBF具有一定的可扩展性。实验结果证明，LGDNBF的分类准确率比传统的基于最小风险的朴素贝叶斯模型有明显的改善。

英文摘要：

This paper analyzed the widely used Naive Bayes filtering model based on risk minimization in spam filtering. With definition of the new risk factor, it put forward a new Naive Bayes filtering model based on line geometry division （LGDNBF）. The new risk factor described risk of decision-making more precisely, and made LGDNBF extendible. The test result demonstrates that LGDNBF has better performance than traditional Naive Bayes filtering model based on risk minimization.