Successive rule refinement towards belief-evidence fusion
College of Computer Studies
This paper discusses successive rule refinement as a method for belief and evidence fusion. The set of “beliefs” is encoded in rule-form (disjunctive normal form) and is the main basis for decision-making. These beliefs in a rule-based system are then confirmed, modified, challenged, or left unsupported by the “evidence” available. Certain new evidences that do not figure in any existing belief are assimilated as new belief. This fusion of belief and evidence is done through successive rule refinement. Evidence is in the form of raw data which have to be converted into rule-form so that they can be integrated with the existing beliefs about the domain. Converting evidence into rule-form is done through a rule extraction system that trains a neural network using the available evidence and extracts rules from the network once it has been sufficiently trained. From the experiments conducted to demonstrate the applicability of the approach, it can be seen that the system’s set of beliefs becomes more and more refined and complete as increasing units of evidence are integrated in it.
Azcarraga, A. P., & Marcos, N. (2001). Successive rule refinement towards belief-evidence fusion. Retrieved from https://animorepository.dlsu.edu.ph/faculty_research/8539
Dempster-Shafer theory; Neural networks (Computer science)