一种改进决策树剪枝算法的研究

决策树归纳方法的剪枝过程是为了消除最终生成的决策树对训练集的过度适应以及减少结点的数量,但最终生成的决策树依然过于庞大。而有些应用对于决策树的精度要求不是很高。通过对剪枝过程加以优化,使得在牺牲少量精度的同时结点的数量大大减少,从而提高生成规则的可理解性。

Effective Simplification of Decision Tree

In decision tree Induction, the purpose of pruning is to prevent over fitting of the training data and reduce the quantity of nodes But the final tree is still too big. The fact is in many applications accuracy is not very important. This paper is to optimize the pruning algorithm, making it a small decrease m accuracy but accompanied by a dramatic reduction in the size of trees.