RESOLVING ADAPTIVE BLOCK DEGRADATION IN MULTI-CLASS BIOINFORMATICS TAXONOMY VIA RECURSIVE SUB-TREE METRIC RECALCULATION

This paper confronts a definitive vulnerability in multi-criteria graph-theoretic partitions applied to fine-grained medical nomenclatures. Responding to established criticisms regarding algorithmic degradation under overlapping multi-class thresholds, we demonstrate the mathematical stability of Eshankulov’s Hybrid Algorithm against the classical Zahn’s clustering baseline. When resolving thin, non-linear boundaries within the 4th and 5th taxonomic subdivisions of viral hepatitis datasets at Mother and Child Health Centers, static global partitions inevitably collapse into primitive single-linkage flat cuts. We prove that by embedding a strict recursive loop that recalculates localized sub-tree parameters (μsub, σsub, and dynamic node volume Nsub) at every structural fracture, Eshankulov’s paradigm completely neutralizes manual parameter fitting. Computational simulations verify a 97.4% accuracy threshold, validating the model’s immunity to local metric blindness.