An Improved D-S Evidence Theory Based on Gray Relational Analysis

Dempster-Shafer (D-S) evidence theory is an uncertain reasoning method which has great application value in data fusion and other areas. However, it can usually lead to counter-intuitive results when fusing the highly conflictive evidences. In this paper, an improved D-S evidence theory based on gray relational analysis is proposed. The proposed method considers the basic probability assignment (BPA) of every evidence as a data sequence, and assigns a weight to each sequence to modify the original evidence model, and then fuses the evidences by D-S combination rule. The new evidence model is more reliable so that the fusion result is more accurate. A numerical example and an experiment on Iris dataset are given to illustrate effectiveness by comparing with other D-S evidence based methods.