A Decision Tree Based On Hierarchy Analysis

At present, people often use decision tree reasoning technique to mine knowledge. The traditional decision tree, which was represented by ID3 proposed by Quinlan in 1986, can solve the classification problem well. But when the class number increases, the produced single decision tree becomes complex and generalization capability decreases. This paper uses the hierarchy analysis method to deal with multi-class problem by producing multi-level decision tree. Compared with the single decision tree, the decision tree based on hierarchy analysis has more advantages when dealing with the multi-class problem.This paper builds decision tree based on hierarchy analysis on the visual C++ software platform and ID3 algorithm. First find out the best combination for the multi-class problem through enumeration. Then analyse and compare the result between ID3 algorithm and hierarchy analysis algorithm from the rule number, the precondition number and the accuracy. At last point out the two algorithms' advantage and disadvantage so as to help us solve the multi-class problem better and establish theory basis for further research of decision tree based on hierarchy analysis.