Influence diagrams are one of the most effective representational tools for decision analysis. However, probabilistic influence diagrams require the availability of probability distributions for all problem's uncertain variables which is not always typical to most real world applications. This paper presents a new approach which adapts these models to real world problems by extending classical influence diagrams within the belief function theory. Hence, we define new graphical decision models called belief influence diagrams which overcome some of the classical influence diagrams limitations such as the necessity of the entire probability distributions. This paper proposes belief evaluation method. It is an adaptation of Shachter method based on arc reversal and nodes removal operations by adding more assumptions specific to belief influence diagrams.
Paper
Full text
Belief influence diagrams
Semantic Scholar · Computer Science · 2014
Abstract
Influence diagrams are one of the most effective representational tools for decision analysis. However, probabilistic influence diagrams require the availability of probability distributions for all problem's uncertain variables which is not always typical to most real world applications. This paper presents a new approach which adapts these models to real world problems by extending classical influence diagrams within the belief function theory. Hence, we define new graphical decision models called belief influence diagrams which overcome some of the classical influence diagrams limitations such as the necessity of the entire probability distributions. This paper proposes belief evaluation method. It is an adaptation of Shachter method based on arc reversal and nodes removal operations by adding more assumptions specific to belief influence diagrams.