Decision tree algorithm based on classification matrix
MA Liang-li · Jisuanji gongcheng yu sheji · 2012
To improve the classification speed and accuracy of the decision tree algorithm,a new program is proposed based on classification matrix.Firstly,the basic theory of the ID3 algorithm is introduced and a classification matrix is defined.Then the variety bias of this algorithm is pointed out,which is proved using the classification matrix.On the basis of the above,a weighting factor is cited to suppress the variety bias of the ID3 algorithm on the premise of a corresponding proof.According to the characteristics of the gain based on the classification matrix,a new decision tree scheme is proposed,aiming to optimize computing speed.Finally,the program is compared with the ID3 algorithm through experiment.Experimental results show that the optimized scheme is obviously better than the original one in performance.