EUVIMEDEUVIMED entdecken
Systematische ÜbersichtMetaanalyse

Machine Learning for Retail Sales and KPI Forecasting: A Systematic Review and Meta-Analysis of Predictive Performance

Songeta Dhar

American Journal of Advanced Technology and Engineering Solutions · 02.08.2025 · 15 (08) · 01

DOI: 10.63125/xan6v087 ↗

Kurz gefasst

Systematische Übersicht in der Zeitschrift American Journal of Advanced Technology and Engineering Solutions, online erschienen am 02.08.2025. Verfasst von Songeta Dhar. Die Arbeit nennt 116 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This study was designed to systematically review and quantitatively synthesize evidence on machine-learning methods for retail sales and key performance indicator forecasting. It focused on comparisons with statistical or other forecasting benchmarks and considered differences in forecast targets, retail settings, model families, prediction horizons, and validation procedures. The proposed review used a systematic search, eligibility screening, structured data extraction, and risk-of-bias assessment. Random-effects meta-analysis was planned for compatible forecast-error outcomes, with subgroup, moderator, and sensitivity analyses used to examine heterogeneity. In the example results, 44 studies were included in the systematic review, 31 contributed to at least one meta-analysis, and 13 were summarized narratively. Across 78 model comparisons, the example pooled error ratio was 0.91 (95% confidence interval [CI] [0.87, 0.96], p = .001), favoring machine-learning models, with substantial between-study variation (I² = 61%). For sales and demand, the RMSE ratio was 0.89 (95% CI [0.82, 0.97], p = .008), while the MAE ratio was 0.92 (95% CI [0.86, 0.99], unadjusted p = .026); the adjusted MAE comparison did not meet the .05 threshold. The example short-horizon estimate favored machine learning (ratio = 0.88, 95% CI [0.82, 0.95], p = .002), whereas the longer-horizon estimate was not statistically significant (ratio = 0.94, 95% CI [0.86, 1.03], p = .170). Sensitivity estimates remained directionally similar after excluding studies with potential temporal leakage. These provisional patterns suggest that comparative performance may vary by metric, retail context, forecast horizon, and study design. Forecast accuracy also does not by itself establish improved inventory availability, lower waste, or higher profitability.

Zitieren

APA

Dhar, S. (2025). Machine Learning for Retail Sales and KPI Forecasting: A Systematic Review and Meta-Analysis of Predictive Performance. American Journal of Advanced Technology and Engineering Solutions, 15(08), 01. https://doi.org/10.63125/xan6v087

Vancouver

Dhar S. Machine Learning for Retail Sales and KPI Forecasting: A Systematic Review and Meta-Analysis of Predictive Performance. American Journal of Advanced Technology and Engineering Solutions. 2025;15(08):01. doi:10.63125/xan6v087

BibTeX
@article{Dhar2025,
  author = {Dhar, Songeta},
  title = {Machine Learning for Retail Sales and KPI Forecasting: A Systematic Review and Meta-Analysis of Predictive Performance},
  journal = {American Journal of Advanced Technology and Engineering Solutions},
  year = {2025},
  volume = {15},
  number = {08},
  pages = {01},
  doi = {10.63125/xan6v087}
}
Kostenlos starten →