Rules extraction method of decision tree based on new conditional entropy
Yuanyuan Ma · Journal of Computer Applications · 2007
The disadvantages of the current information entropy for estimating decision ability were analyzed deeply.To eliminate the limitations,a new conditional entropy was defined.The attribute selection metric of traditional heuristic algorithm was modified,so the new improved significance of an attribute was proposed.Finally,a heuristic algorithm for rules extraction of decision tree was designed.This reduction method does not need attribute reduction before extracting decision rules,and its computation is direct and efficient,and its time complexity is less than the others.The experiment and comparison show that the algorithm provides more precise and simple decision rules.