Improvement of attribute selection criterion of decision trees
Liu Yu-xun · Computer Engineering and Applications Journal · 2010
The decision tree algorithm is a research hotspot in the field of data mining,which is usually used form classifiers and prediction models.In practice,it is widely used.This paper focuses on classical ID3 algorithm,analyzes its advantages and disadvantages,combines with Taylor and Maclaurin formula,and then puts forward a new attribute selection criterion of decision trees.This modified algorithm improves the classification accuracy,reduces the generation time of decision trees,and shortens the computational cost by the calculation of the simplified information entropy.Experimental results show effectiveness and correctness of the improved algorithm.