EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

AI-Powered Decision Agents for Fashion Merchandising: An Interpretable Multi-Agent Machine Learning Framework for Trend Forecasting and Product Planning

Mahabub Rahman

International Journal of Applied Data Analytics · 2026

DOI: 10.51137/wrp.ijada.977 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Applied Data Analytics, 2026. Verfasst von Mahabub Rahman. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

In the contemporary fast-fashion ecosystem, retailers face severe volatility driven by rapid trend cycles and the cold-start problem of new product launches. Traditional univariate models fail to capture visual aesthetics and consumer sentiment, while state-of-the-art deep learning architectures operate as “black boxes” that lack the transparency required for high-stakes inventory decisions. To resolve these challenges, this study presents an interpretable Multi-Agent Machine Learning (MAML) framework that integrates predictive analytics with strategic product planning. The proposed methodology harmonizes three large-scale multimodal datasets, Visuelle 2.0, DeepFashion2, and Amazon Reviews 2023, and orchestrates four specialized agents: Trend Sensing, Demand Forecasting, Product Planning, and Explanation. Empirical validation demonstrates that this multimodal approach achieves a 43.1% relative reduction in Weighted Absolute Percentage Error (WAPE) over univariate ARIMA (reducing WAPE from 32.5% to 18.5%) and a 23.6% reduction over standard LSTM baselines. In cold-start scenarios across the first four post-launch weeks, synthesizing visual and textual signals captures initial demand spikes with 82% directional accuracy. Furthermore, optimizing procurement depth through constrained optimization yields a 19% reduction in end-of-season markdown risk. By embedding SHAP-based feature attribution, the Explanation Agent provides human-readable rationales, bridging the trust gap between autonomous AI systems and executive merchandising practitioners.

Zitieren

APA

Rahman, M. (2026). AI-Powered Decision Agents for Fashion Merchandising: An Interpretable Multi-Agent Machine Learning Framework for Trend Forecasting and Product Planning. International Journal of Applied Data Analytics. https://doi.org/10.51137/wrp.ijada.977

Vancouver

Rahman M. AI-Powered Decision Agents for Fashion Merchandising: An Interpretable Multi-Agent Machine Learning Framework for Trend Forecasting and Product Planning. International Journal of Applied Data Analytics. 2026. doi:10.51137/wrp.ijada.977

BibTeX
@article{Rahman2026,
  author = {Rahman, Mahabub},
  title = {AI-Powered Decision Agents for Fashion Merchandising: An Interpretable Multi-Agent Machine Learning Framework for Trend Forecasting and Product Planning},
  journal = {International Journal of Applied Data Analytics},
  year = {2026},
  doi = {10.51137/wrp.ijada.977}
}
Kostenlos starten →