基于群体智能的选择性决策树分类器集成

Swarm Intelligence-Based Selective Ensemble with Decision Trees Classifiers

尽管选择性集成方法的研究和应用已取得了不少重要成果,然而其实现方法计算复杂度高、效率低仍是应用该方法的一个瓶颈。为此,提出了一种新的高速收敛的选择性集成方法。该方法使用C4.5决策树分类器作为基学习器,利用高速收敛的群体智能算法来寻找最优集成模型,并在UCI数据库的多值分类数据集上进行了实验。实验结果表明,该方法计算效率高,其精度和稳定性比Bagging方法都要高,可以成为一种高效的选择性集成的实现方法。

Although a good many important results have been achieved about the research of selective ensemble approach and its application, it remains a computational bottleneck that the implementstion of selective ensemble approach costs too much time to find an optimal ensemble. Therefore,a quickly convergent version of selective ensemble algorithm is presented. This algorithm uses convergent SI （swarm intelligence） to find the optimal ensemble with using the C4.5 decision trees classifiers as based learners. Meanwhile,experiments are carried out on UCI data sets. The computer experiments demonstrate that the proposed algorithm achieves high speed, and its accuracy and stability are both higher than Bagging algorithm. It can become a high efficient selective ensemble algorithm.