Evaluation Information Fusion of Scientific Research Project Based on Evidential Reasoning Approach Under Two-Dimensional Frames of Discernment

• Main Authors: Weidong Zhu, Shaorong Li, Quan Ku, Chao Zhang
• Format: Article
• Language: English
• Published: IEEE 2020-01-01
• Series: IEEE Access
• Subjects: Evaluation information fusion; scientific research project; two-dimensional frame; evidential reasoning
• Online Access: https://ieeexplore.ieee.org/document/8949360/

Owing to the complexity and uncertainty of scientific research projects, government-funded institutions often invite multiple peer experts to evaluate the projects. How to aggregate the evaluation opinions of multiple peer experts is an important issue. With consideration of the experts' knowledge background and historical evaluation performance, this paper proposes an evidential reasoning approach under two-dimensional frames of discernment to aggregate the evaluation opinions of multiple experts. Firstly, two-dimensional frames of discernment are constructed. The I-dimensional frame of discernment describes experts' evaluation opinions, and the II-dimensional frame of discernment describes experts' knowledge background and historical evaluation performance. Experts' evaluation opinions and characteristic information are transformed into pieces of evidence, which could be expressed by their respective belief distributions. Secondly, the II-dimensional evidence information is used to generate the evidence correction factor to reflect experts' reliabilities. Finally, the experts' evaluation opinions are corrected using the evidence correction factors. Thus, the corrected experts' evaluation opinions contain the experts' characteristic information. In addition, the evidential reasoning approach is used to aggregate the evaluation opinions of multiple experts to complete the selection of the scientific research projects. An illustrative example on aggregating the evaluation opinions of the National Natural Science Foundation of China is provided to demonstrate the applicability and validity of the proposed method. Experimental results show that the proposed method can enhance the description of uncertainty in experts' evaluation opinions, reflect the qualities of the experts' evaluation opinions, and make the aggregation results of multiple experts' evaluation opinions more reasonable and accurate.