一种基于差异思想的选择性Bagging Trees集成算法研究

The Study of A Selective Bagging Trees Ensemble Based on Diversity Thought

本文通过对鸢尾花数据的研究,提出了一种基于分类器的分类效果差异而进行快速选择的一种改进的Bagging Trees集成算法。并通过同其他统计机器学习方法,如:CART、Bagging Trees、Random Forest以及目前流行的基于遗传算法的选择性集成算法GASEN等比较得出,该算法对于分类问题而言,具有较高的准确率,而且与GASEN算法相比,运行的效率也得到了较大的提高。

In this paper, based on a case study of iris dataset, it draws a new ensemble algorithm, a selective bagging trees ensemble based on diversity of different classifiers. And contrasted with other statistical machine learning methods, such as, CART, bagging trees, random forest and the current prevalent selective ensemble based on genetic algorithm, GASEN, this new algorithm proposed in this paper has higher accuracy, and also costs much little time than GASEN algorithm and improves efficiency when it is used in the problems of classification.