Abstract: The methods of pruning have great influence on the effect of the decision tree. By researching on the pruning method based on misclassification, this paper introduces the conception of condition misclassification and improves the standard of pruning, it proposes the conditional misclassification pruning method for decision tree optimization and applies it in C4.5 algorithm. Experimental result shows that the condition misclassification pruning can avoid over pruned problem and nonenough pruned problem to some extent and improve the accuracy rate of classification.

Key words: decision tree, misclassification pruning, condition misclassification

摘要： 在建立决策树分类模型时，剪枝的方法直接影响分类器的分类效果。通过研究基于误差率的剪枝算法，引入条件误差的概念，改进剪枝标准的评估方法，针对决策树的模型进行优化，提出条件误差剪枝方法，并将其应用于C4.5算法中。实验结果表明，条件误差剪枝方法有效地解决剪枝不充分和过剪枝的情况，在一定程度上提高了准确率。

关键词: 决策树, 误分类剪枝, 条件误分类

CLC Number: 

• TP311.52