【摘要】 TAN分类器以良好的分类性能而著称,但分类器本身和归纳学习算法并不具有处理丢失数据的能力,而现有的用于分类技术中丢失数据处理的方法在可靠性方面均不同程度地存在一些缺陷.本文针对问题,结合TAN结构和Gibbs sampling进行具有丢失数据的分类器迭代学习,在迭代中,TAN结构学习、参数学习和丢失数据修复交替进行,随着迭代的收敛,最终将得到TAN分类器,同时丢失的数据也得到修复.更多还原

【Abstract】 TAN classifier is well-known due to its outstanding performance.At present,neither TAN classifier nor inductive learning arithmetic has the function of processing missing data.Existing methods working on missing data in classification technology need to be improved in reliability.In this paper,an iterative method of learning TAN classifiers with missing data is presented by combining TAN structure with Gibbs sampling.In iteration,building TAN structure,learning TAN parameters and revising unobserved data are alternately done.When iteration convergence TAN calssifier with missing data will be obtained.And missing data can also be repaired.更多还原