基于Boosting算法的C5.0决策树不平衡数据分类算法

Boosting-based C5.0 Decision Tree Algorithm for Imbalanced Data Classification Problem

为了改进不平衡数据的分类性能,提出一种可自动确定迭代参数trail值的集成C5.0决策树算法.首先,算法引入boosting集成框架到C5.0决策树算法中,从而生成新的集成分类器;其次,算法使用网格搜索法在一定范围内自动确定trail参数的值.实验结果表明,该算法在不平衡数据上的分类性能指标G-mean和MCC上具有优势.

To improve the impact of imbalanced data on classification performance,an ensemble C5.0 decision tree algorithm which can automatically determine the value of the iterative parameter trail is proposed.Firstly,the boosting ensemble framework is introduced into the C5.0 decision tree algorithm to generate a new ensemble algorithm B-C5.0.Secondly,the grid search algorithm is used to search the optimal trail parameter.The experimental results show that the algorithm has advantages in the classification performance index G-mean and MCC on imbalanced data.