关键词: 集成协同训练, 旋转森林, 差异性, 特征提取, 旋转森林的协同训练方法（ROFCO）

Abstract: Ensemble co-training is a kind of semi-supervised learning method which combines ensemble learning and co-training.Rotation forest is a kind of ensemble learning which generates classifier ensembles based on feature extraction.A novel ensemble co-training algorithm which named ROFCO is proposed which focuses on using unlabeled data to improve the diversity between base classifiers and feature extraction effect.The base classifiers will maintain or decrease the generalization error，while maintaining or even improving diversity between them.Experiments on UCI data set and the benchmark data set demonstrate that compared with other similar algorithms ROFCO could get much better performance.

Key words: ensemble co-training, rotation forest, diversity, feature extraction, Rotation Forest Co-training（ROFCO）