Journal of King Saud University Computer and Information Sciences Компьютерные науки (общие)
Conflict inherits the frame: Bin geometry and BPA construction in evidential fusion of quantised continuous evidence
Аннотация
Abstract Dempster-Shafer fusion of continuous measurements requires a frame of discernment, normally built by binning the measurement into intervals. This step decides what the conflict mass measures. When the intervals differ in width, Dempster’s conflict K is confounded with the position of the evidence: under any fixed-bandwidth basic probability assignment (BPA), two identical readings conflict more near a narrow interval than near a wide one. On our data the confound reverses the apparent conflict-severity relationship. An excess-conflict statistic removes it at negligible cost and Jousselme’s distance is free of it; on such a frame both reduce to a monotone reparametrisation of the plain difference between the readings, so both are diagnostics, not new signals. A second failure mode is the encoding: a point mass on the interval containing each reading makes K a two-valued disagreement indicator (r = 0.998 across 13,221 records), and with singleton focal elements Dempster’s rule becomes a discounted naive-Bayes product over intervals, so the combination rule barely matters. We demonstrate both on 37,975 CVE-2024 records from three severity databases, over a 4 × 6 × 4 grid of BPA constructions, rules and measures under nested cross-validation; 13,221 dually assessed pairs with 70 confirmed-exploited positives are the analytical sample, the third source entering only descriptively. Fusion yields an efficacy null result reported with its power; the average-precision gain is directionally consistent over twenty partitions but not significant under every resampling scheme; and the agreement-exploitation relationship holds on the records whose scores predate their listing.