Dempster–Shafer evidence theory (evidence theory) has been widely used as an efficient method for dealing with uncertainty. In evidence theory, Dempster's rule is the most well-known evidence combination method but it does not work well when the evidence is in high conflict. To improve the performance of combining conflicting evidence, an original and novel evidence combination method is presented based on the Pearson correlation coefficient and weighted graph. The proposed method can correctly recognize the alternative situation with a high accuracy. Besides, the convergence performance of this method is better when compared with other combination rules. In addition, the weighted graph generated by the proposed method can directly represent the relationship between different evidence, which can help researchers estimate the reliability of different body of evidence. Our experimental results indicate the advantages of our proposed evidence combination rule over existing methods, and the results are analyzed and discussed.

Information quality has received extensive attention recently. Yager and Petry proposed an information quality suitable for the framework of probability theory, and proposed a method of fusing multisource information, which can improve the information quality required for decision-making. Then, Bouhamed et al. extended information quality to the theory of possibility. However, the basic probability assignment (BPA) in evidence theory can deal with uncertainty more effectively. Therefore, this work provides a companion paper that makes the method applicable to evidence theory. This method uses vector notation to represent B P A. A fusion method is designed to select the best quality subset based on two factors: information quality and source credibility function, and using the score function to verify the quality of each subset. Finally, a numerical example details the eight steps of the method, and uses the Iris data set and banknote authentication data set to illustrate the application of the method in pattern recognition.