<![CDATA[This research aims to determine which scientific claims remain supportable when an archived YOLOv8-fuzzy traffic prototype lacks essential experimental artifacts and to demonstrate a preliminary two-axis scheme for separating artifact status from validation status. The case comprised four unsynchronized road-video files, retained detector graphics and a count matrix, four snapshot-load/displayed-time pairs, and one interface state. The audit inventoried artifact provenance; classified evidence as unavailable, retained, deterministically recomputable, or assumption-dependently reconstructed; recalculated classwise and pooled precision, recall, and F1 from 1,853 true-class instances; and used uniformly sampled, constraint-filtered Mamdani parameter configurations to assess finite-data parameter non-uniqueness. Evaluation used recomputed detector metrics, internal consistency with retained timing pairs, and response-envelope widths over unobserved loads. Micro-precision, micro-recall, and micro-F1 were 0.905, 0.934, and 0.920, respectively, although cars and motorcycles contributed 99.6% of the available support. Three thousand accepted surrogate configurations matched the four retained timing pairs by design but diverged elsewhere, producing maximum envelope widths of 10.22-14.66 s. The archive therefore supported limited metric recomputation and interface-consistency analysis, but not detector generalization, recovery of the original controller, or operational traffic-control effectiveness. The scheme is intended for researchers, reviewers, and repository curators who must evaluate incomplete legacy intelligent-transportation or other soft-computing prototypes without overstating retained evidence.]]>
Copyrights © 2026
