基于熵的决策树分枝合并算法

目前,基于逻辑的示例学习算法主要分两大类,决策树算法和基于规则的算法,前者以ＩＤ３为代表,ＩＤ３使用“信息熵”作启发式得出较小的决策树,但ＩＤ３算法只注意到减少树的深度,忽视树的宽度,本文给出了一种决策树分枝合并算法。可减少决策树的宽度,从而得出比ＩＤ３更好的结果。

Algorithm for Merging of Branches in Decision Tree Induction

At present, there are two types of algorithms for leaming from example (splitting algorithm, agglomerative algorithm). Splitting algorithms represent knowledge by decision tree. ID3 is the most famous among them. Entropy. is utilized to generate .smaller decision tree. (fewer nodes , fewer leaves). But ID3 lays emphasis on reducing .the depth of tree. and doesn'tgive attention to " the width of tree" . The decision tree resulted soroetimes is larger. So thealgorithm EM ID is developed. lt can generate the decision tree smaller than ID. in inductionby merging some branches in decision tree.