A Maximum contribution method for classification based on information theory
Lin Ke-ming, Xue Yong-sheng, Wen Juan · 2008
Inductive learning for classification based on information theory is one of the important topics in data mining. We here propose an Maximum contribution method for classification based on information theory. According to the theory of channel transmission in information theory, the definition contribution is developed based on probability distribution of classified space, probability transfer matrices of classified space and feature space and mutual information, then entities is classified by the Maximum contribution method. Finally the empirical test and analyses prove the feasibility of the method.