摘要
首先介绍了一种朴素贝叶斯增量分类模型,然后提出了一种新的序列学习算法以弥补其学习序列中存在的不足训练实例的先验知识得不到充分利用,测试实例的完备性对分类的影响在学习过程中得不到体现等。该算法引入一个分类损失权重系数λ,用于计算分类损失大小。引入该系数的作用在于充分利用先验知识对分类器进行了优化;通过选择合理的学习序列强化了较完备数据对分类的积极影响,弱化了噪音数据的消极影响,从而提高分类精度;弥补了独立性假设在实际问题中的不足等。
This paper first introduces an incremental classification model of nave Bayes,then puts forward a new se-quence learning algorithm to make up the deficiency of such models,such as prior knowledge of training instances could not be fully made use of,influence of the maturity of the testing instances upon classification couldn't be visual-ized in the learning process.This algorithm introduces a classifying loss weight coefficient λfor each training instance in order to calculate the total classifying loss.After introducing the coefficient ,the classifier is optimized by fully utiliz-ing the prior knowledge;by means of choosing reasonable learning sequence,positive influence of the maturer data on classification is strengthened and negative influence of the noise data is weakened and as a result,classification preci-sion is improved;deficiency of the independency assumption in practical operation is also made up.
出处
《计算机工程与应用》
CSCD
北大核心
2004年第14期57-59,共3页
Computer Engineering and Applications
基金
安徽省自然科学基金(编号:03042305)资助
