A new basic probability assignment generation and combination method for conflict data fusion in the evidence theory

Dempster-Shafer evidence theory is an effective method to deal with information fusion. However, how to deal with the fusion paradoxes while using the Dempster's combination rule is still an open issue. To address this issue, a new basic probability assignment (BPA) generation method based on the cosine similarity and the belief entropy was proposed in this paper. Firstly, Mahalanobis distance was used to measure the similarity between the test sample and BPA of each focal element in the frame of discernment. Then, cosine similarity and belief entropy were used respectively to measure the reliability and uncertainty of each BPA to make adjustments and generate a standard BPA. Finally, Dempster's combination rule was used for the fusion of new BPAs. Numerical examples were used to prove the effectiveness of the proposed method in solving the classical fusion paradoxes. Besides, the accuracy rates of the classification experiments on datasets were also calculated to verify the rationality and efficiency of the proposed method.

Automated summary

This paper proposed a new method for generating basic probability assignments (BPA) based on cosine similarity and belief entropy to address the fusion paradoxes when using Dempster's combination rule. The method measures the similarity between the test sample and BPA of each focal element using Mahalanobis distance, and then uses cosine similarity and belief entropy to adjust and generate a standard BPA. The effectiveness of the proposed method in solving classical fusion paradoxes and its rationality and efficiency were verified through numerical examples and classification experiments on datasets.