基于模型决策树的AdaBoost算法

AdaBoost算法是一种将多个基学习器通过合理策略结合生成强学习器的集成算法，其性能取决于基学习器的准确率和多样性。但弱学习器分类精度不高往往也导致了最终强分类器性能较差，因此进一步为了提高算法的分类精确率，本文提出一种MDTAda模型，首先利用基尼指数迭代构造一棵不完全决策树，然后在决策树的非纯伪叶结点上添加简单分类器，生成MDT(模型决策树),将MDT作为AdaBoost算法的基分类器，加权平均生成强分类器。在标准数据集上的实验表明，相比传统的AdaBoost算法，本文提出的算法拥有更好的泛化性能和更优的间隔分布，且在与AdaBoost算法达到相同精度时所需迭代次数更少。

AdaBoost algorithm based on model decision tree

The AdaBoost algorithm is an ensemble algorithm that combines multiple base learners through reasonable strategies to generate a strong learner. Its performance depends on the accuracy and diversity of the base learners. However, the poor classification accuracy of weak learners often leads to poor performance of the final strong classifier. Therefore, in order to further improve the classification accuracy of the algorithm, this paper proposes an MDTAda model, which first uses Gini index to iteratively construct an incomplete decision tree. Then add a simple classifier to the non-pure pseudo-leaf nodes of the decision tree to generate MDT(model decision tree), use MDT as the base classifier of AdaBoost algorithm, and weighted average to generate a strong classifier. Experiments on standard data sets show that compared with the traditional AdaBoost algorithm, the algorithm proposed in this paper has better generalization performance and better margin distribution, and requires fewer iterations to achieve the same accuracy as AdaBoost algorithm.