Validating Selective Prediction Under Source Uncertainty in Small Enterprise Surveys
International Journal of Advanced Computer Science and Applications · 2026 · 17 (10)
DOI: 10.14569/ijacsa.2026.0171056 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Advanced Computer Science and Applications, 2026. Verfasst von Suyono Thamrin und 2 weiteren Autor:innen.
Abstract (Originalsprache)
Survey-based prediction is sensitive to unresolved source versions, repeated predictor profiles, and abstention rules. We evaluate these choices jointly in two Indonesian culinary-enterprise files, containing 206 aligned positions and 301 recorded rows, respectively. Each file has 177 predictor profiles, without verified respondent identities. Five fixed baselines are compared over 50 overlapping partitions, supplemented by training-only nested tuning, hierarchical conformal calibration, profile-stratified diagnostics, and explicit review-cost scenarios. In the second file, random-forest mean absolute error increases from 0.1966 under row splitting to 0.2814 under profile-disjoint splitting; ridge achieves 0.2369. A limited tuning sensitivity preserves this ordering. In the first file, simultaneous coverage of recorded profile outcomes and versions is 90.00 percent for profile-version calibration and 87.28 percent for row-midpoint calibration. Approximate dependence-adjusted uncertainty intervals include zero for the paired coverage gain, so superiority is not established. Aggregate coverage also conceals poor performance in some response and repetition strata. Exploratory abstention analyses reduce retained error but can lower coverage or require infinite intervals. Additional ordinal simulations expose failure under noise shift and unrecorded versions, while apparent coverage under misalignment does not authenticate the source. The contribution is an auditable evaluation workflow using established methods, with explicit coverage events, implementation choices, and decision-cost assumptions. Its transferable value is diagnostic; learner rankings, innovation effects, population validity, and operational benefits cannot be generalized from these files.
Zitieren
APA
Thamrin, S., Capnary, M. C., & Agung, I. (2026). Validating Selective Prediction Under Source Uncertainty in Small Enterprise Surveys. International Journal of Advanced Computer Science and Applications, 17(10). https://doi.org/10.14569/ijacsa.2026.0171056
Vancouver
Thamrin S, Capnary MC, Agung I. Validating Selective Prediction Under Source Uncertainty in Small Enterprise Surveys. International Journal of Advanced Computer Science and Applications. 2026;17(10). doi:10.14569/ijacsa.2026.0171056
BibTeX
@article{Thamrin2026,
author = {Thamrin, Suyono and Capnary, M. Calvin and Agung, Iskandar},
title = {Validating Selective Prediction Under Source Uncertainty in Small Enterprise Surveys},
journal = {International Journal of Advanced Computer Science and Applications},
year = {2026},
volume = {17},
number = {10},
doi = {10.14569/ijacsa.2026.0171056}
}
EUVIMED