Investigating Another Weighted Conflict Redistribution Rule of Combination
Dan Harris, Jean Dezert · 2024
This paper investigates another global fusion (i.e., N-dimensional) rule of combination (NRoC) called the Weighted Conflict Redistribution #2 (WCR2), which fuses disparate sources of information simultaneously by making use of tensor geometry. WCR2 is the second NRoC of its kind, second to Harris-Lovassy NRoC, which is WCR1. These WCRs address two important fusion problems within the Dempster-Shafer framework: first, it addresses a perceived shortfall of Dempster’s rule of combination (DR)—which is its numerical instability given a violation of its tacit assumption that expert forecasted probabilities are reliable—by accounting for less than perfectly reliable probabilities; second, it avoids issues arising from lacking the associative property by fusing information simultaneously. In order to evaluate WCR2, a Monte Carlo simulation was run to generate performance metrics and compare WCR2 against WCR1 and DR under various conditions. It was shown that WCR2 produced more reliable fused probabilities than the other two fusion methods, while keeping fused decision accuracy within family of the other two fusion methods.