一种多分类器协同的半监督分类算法SSC_MCC

摘要

为了提高半监督分类性能，提出了一种多分类器协同的半监督分类算法SSC_MCC 。算法采用双层结构集成，使用多条件判断挖掘未标记样本信息，扩充有标记样本。第一层中，采用三分类器协同投票一致策略实现对未标记样本进行标记，第二层中采用基于正确分类率的分类器加权投票决策标记未标记样本，扩充有标记样本，用最终生成的有标记样本训练分类器，实现半监督分类。最后，使用UCI数据集模拟半监督实验，结果表明SSC_MCCL较好地提高了半监督分类性能。%In order to improve the performance of semi⁃supervised learning,a kind of semi⁃supervised classification algorithm based multiple classifier cooperation(SSC_MCC)was proposed. The algorithm was composed of double layer structure integration,using multi condition judge to dig the unlabeled samples information and expand the labeled samples. In the first layer,collaborative voting strategy using three classifiers was to label the unknown sample. In the second layer,the weighted voting decision strategy based on correct classification rate was used to label the unknown sample and expand the labeled sample,using the generated samples to train the classifier and come true the semi⁃supervised classification. Finally,experiment was carried out based on the UCI data set. The results showed that SSC_MCC can improve the classification performance of semi⁃supervised learning.