关键词: 集成学习, 选择性集成, Q统计量

Abstract: Diversity among the individual learners is deemed to be a key issue in ensemble learning.Popular ensemble learning algorithms such that Bagging adopts re-sampling technology to produce the diversity of the individual learners.Selective ensemble chooses individual learners which are a part of the original learners to ensemble.The result shows that it is better than the original ensemble.However，how to choose learners is a problem.Using Q statistic to measure the diversity of a pair of learners，a new selective ensemble learning method for decision tree is proposed.Compared with the C4.5 and Bagging method，it works better.

Key words: ensemble learning, selective ensemble, Q statistic