Decision-theoretic rough set model of multi-source decision systems

Through the complementary integration of information from different sources, information fusion can improve the decision-making process in the increasing uncertain environments. How to make full use of the information from various sources to make decisions is a key problem in multi-source decision-theoretic. The more accurate and comprehensive information is, the easier the decision will be. Thus the uncertainty of decision-making is an objective criterion for evaluating the fusion effect. Therefore, three kinds of multi-source decision methods are proposed based on considering the uncertainty of decision-making process, which are the conditional entropy multi-source decision (CE-MSD) method, the decision support degree multi-source decision (DS-MSD) method and the mean multi-source decision (M-MSD) method. The CE-MSD method based on taking into account the uncertainty of each condition attribute for decisions aims to select the most reliable source for each attribute according to the conditional entropy, and then make final decisions under a new restructuring decision table. The DS-MSD method proposed by considering the uncertainty of condition attribute set for decisions aims to make the final decision through the decision support degree of all the sources to each object. The M-MSD method and the approximate precision index are introduced as reference standards to measure the effectiveness of CE-MSD and DS-MSD in the multi-source decision system. Meanwhile, three corresponding algorithms are designed to verify the effectiveness and feasibility of the proposed decision methods. Finally, in order to verify the validity of methods, approximation accuracies of CE-MSD, DS-MSD and M-MSD are compared in multi-source decision systems which are generated by adding Gauss noise and random noise to Data set downloaded from UCI. In sum, the decision theory of multi-source decision systems is a generalization of the decision-theoretic rough set, which is worthy of further research.