Abstract: Ensemble of complete random trees, i.e. decision trees without any split selection, has high performance. This paper investigates whether the complete random learning scheme can by applied to other types of base learners. It realizes complete random decision stump and complete random rule algorithms, analyzes why complete scheme work. Experiments show that complete random scheme works for different types of base learners.

Key words: Machine learning, Ensemble learning, Complete random scheme

摘要： 以完全随机树(不包含属性选择过程的决策树)作为基学习器的集成，具有很好的性能。该文探讨了完全随机学习策略推广情况，实现了完全随机决策树桩算法和完全随机规则算法，分析有效的原因。实验表明，性能良好的完全随机算法，易于被许多初学者所掌握。

关键词: 机器学习, 集成学习, 完全随机策略

CLC Number: 

• TP181