A Research On Crisp Decision Tree Induction Algorithm Based On Classification Ambiguity

Decision tree algorithm can be divided into two types: crisp decision tree and fuzzy decision tree. Typically, the fuzzy decision tree algorithm is an improvement of the crisp decision tree algorithm, and is an extension of the crisp decision tree algorithm. From the opposite point of view, this article presents a new crisp decision tree algorithm by improving the fuzzy decision tree algorithm proposed by Yuan.The paper first introduces the theory of decision tree, gives several commonly used crisp decision tree algorithms and fuzzy decision tree algorithms, and then, proposes a crisp decision tree algorithm based on classification ambiguity by improving Yuan's fuzzy decision tree algorithm. Finally, we compares our algorithm with other crisp decision tree algorithm: First, we compare our crisp decision tree algorithm with ID3 algorithm on the discrete databases, and compare our crisp decision tree algorithm with C4.5 algorithm on the continuous databases, the results show that on the condition of reaching classification accuracy of ID3 and C4.5, our algorithm can induce a smaller tree than ID3 and C4.5; and then, we compare heuristic function used in our algorithm with that based on information entropy and point out that the heuristic function we used has different characteristics from the information entropy heuristic function.