“School of Cognitive Sciences”

Back to Papers Home
Back to Papers of School of Cognitive Sciences

Paper   IPM / Cognitive Sciences / 15237
School of Cognitive Sciences
  Title:   How to decide when the sources of evidence are unreliable: A multi-criteria discounting approach in the Dempster-Shafer theory
  Author(s): 
1.  A. Sarabi-Jamab
2.  B. Araabi
  Status:   Published
  Journal: Information Sciences
  Vol.:  448-449
  Year:  2018
  Pages:   233-248
  Supported by:  IPM
  Abstract:
Dempster-Shafer theory manages uncertain information and offers a useful fusion tool for decision-making. However, the quality of combining is affected by conflicting information, especially when the sources of evidence are unreliable. A number of methods for discounting unreliable sources of evidence have been proposed. In these approaches, the estimation of the discounting factors is crucial, mainly when prior knowledge is unavailable. Most discounting methods assume that the distance between bodies of evidence is the only factor for conflict estimation. However, considering just one factor is not reliable enough to assess the reliability degrees, since the confusion, disparity, and imperfection of information provided by the sources of evidence are due to different types of conflict. In this paper, we proposed a multi-criteria method to estimate the reliability factors. We defined a method to select a comprehensive set of criteria, which each criterion takes into account the uncertainty of each source of evidence. Then, we proposed an aggregation method to estimate the reliability degree. Finally, we extended our proposed reliability degrees in a sequential discounting approach. We verified our suggested discounting method through two designed experiments. The results showed that our approach was efficient in combining the unreliable sources of evidence. Moreover, our proposed reliability degree was more robust in outliers than the previous methods.

Download TeX format
back to top
scroll left or right