摘要
朴素贝叶斯(Nave Bayes,NB)分类方法是一种简单而有效的概率分类方法,但是贝叶斯算法存在训练集数据不完备这个缺陷。传统的贝叶斯分类方法在有新的训练样本加入时,需要重新学习已经学习过的样本,耗费大量时间。为此引入增量学习算法,算法在已有的分类器的基础上,自主选择学习新的文本来修正分类器。本文给出词频加权朴素贝叶斯分类增量学习算法思想及其具体算法,并对算法给予证明。通过算法分析可知,相比无增量学习的贝叶斯分类,本算法额外的空间复杂度与时间复杂度都在可接受范围。
Naive Bayes(NB) Classifier is a simple and effective classification method based on probability theory. But this algorithm has some lack and limitation, especially the non - maturity training - data collection. Traditional NB classification must costs a lot of time to learn all samples again when new sample added. So a concept of incremental learning algorithm is put forward. The algorithm based on exist classifier, using new information from new sample to modify the classifier. We introduce the main meaning of incremental learning algorithm based on Weighted Naive Bayes. And we give the very algorithm and prove it. Through the analysis of the algorithm, we get the result that Comparing to non - Incremental learning algorithm, the additional space complexity and time complexity of incremental learning algorithm are both in the acceptable range.
出处
《微计算机应用》
2008年第6期107-112,共6页
Microcomputer Applications
关键词
加权朴素贝叶斯
增量学习
分类算法
类置信度
Weighted Naive Bayes, Incremental Learning, Classification Algorithm, Class Confidence