结合主动学习的半监督贝叶斯网络多类标分类-信号与信息处理专业论文.docxVIP

II II Abstract The problem of multi-label is common in the real world..In these multi-label problems,each sample of training set is corresponding to composed of HYPERLINK /dict_result.aspx?searchword=%e4%b8%80%e4%b8%aa%e6%88%96%e5%a4%9a%e4%b8%aaamp;tjType=sentenceamp;styleamp;t=one%2Bor%2Bseveral one or several class-label.Multi-label learning mission is forecast class-label for each unknown sample when the size of class-label set is unknown.The problem of multi-label classification is evidently different from traditional classification which each sample only belongs to one class.Single-label algorithm can’t directly used for processing these multi-label data. Traditional multi-label classification algorithm is improved on the base of common single-label algorithm.Classification results of these algorithms are satisfactory to some extent,but all depend on the training data set need for machine learning. The drastical cost in preparing enough labeled samples is too expensive,so how to use large number of unlabeled samples to find the little quality samples to get the effective classifier by training is a troublesome problem in Multi-Label field. Traditional multi-label classification algorithm requires a large number of labeled data in order to model problem efficiently. This paper proposed a multi-label classification algorithm employing Gaussian fields and Bayesian network to combine active learning and semi-supervised learning. First,Multi-label data often has high dimensional feature attributes, in this condition, a dimension-reducing algorithm using LLE was presented to build several Bayesian network multi-label algorithms. Then,on the basis of Gaussian random field model, a algorithm which is named ML-ASGB of combining active learning and semi-supervised of Bayesian Networks on Multi-Label Classification is proposed. This algorithm enables Bayesian network multi-label classification to select a small number of high quality sample human label and model problem