摘要
满足条件独立性假设时,朴素贝叶斯分类器理论上比其它分类方法具有更高的分类正确率,但该假设在许多实际情况中并不成立,针对这一问题,提出了一种基于因子分析的朴素贝叶斯分类模型FA-NBC,并将其应用于边坡的稳定性识别。为了保证朴素贝叶斯分类器结构上的简单性,FA-NBC模型以方差贡献为依据构建新的属性集,新属性集包含原属性集的大部分信息且满足条件独立性假设。UCI数据集上的实验结果证明了FA-NBC模型的有效性。
Naive Bayesian classifier （NB） is popular for its simplicity and effectiveness. However, the accuracy of NB is affected when the conditional independence assumption is violated. A new algorithm based on factor analysis, FA-NBC, is proposed to retain the structure strength of NB while reducing error by alleviating the attribute interdependence problem. Then the classifier is applied to slope recognition. New independent attribute set which includes most of the information of the original property set is built based on variance to ensure the structural simplicity of naive Bayesian classifier. Experimental results on UCI data sets prove the validity of the model.
出处
《计算机工程与设计》
CSCD
北大核心
2011年第11期3828-3831,共4页
Computer Engineering and Design
基金
国家自然科学基金项目(61070033)
广东省自然科学基金重点项目(9251009001000005)
广东省科技计划基金项目(2010B050400011
2008B080701005)
广东省哲学社会科学规划"十一五"规划基金项目(08O-01)
关键词
朴素贝叶斯分类
条件独立
属性选择
因子分析
边坡稳定性识别
naive Bayesian classifier
conditional independence
attribute selection
factor analysis
slope recognition
