Table of Contents: :: Library Catalog

Saved in:

Bibliographic Details
Main Authors:	Shambit, Md Kawsar Afjal, Fuad, Md Shamimun, Sany, Md Zahidul Islam
Format:	Recurso digital
Language:	English
Published:	Zenodo 2025
Online Access:	https://doi.org/10.5281/zenodo.17412895
Tags:	Add Tag No Tags, Be the first to tag this record!

Table of Contents:

Abstract—This paper presents a hybrid machine learning framework that addresses scalability and accuracy challenges in retail inventory management by integrating real-time demand forecasting with anomaly detection, validated using Walmart’s historical sales data. Traditional forecasting methods face a critical trade-off: maintaining separate models for each product category is computationally intensive, while generalized models often underperform for diverse items, resulting in stockouts or overstocking. To overcome this limitation, we introduce a department-level aggregation strategy that balances specificity and generalization. The proposed hybrid approach combines ARIMA for baseline statistical forecasting, cubic spline interpolation to capture non-linear trends, and neural networks to model complex feature interactions. The framework continuously refines predictions using real-time sales streams and employs residual analysis with adaptive thresholding to detect anomalies such as abrupt demand surges or supply disruptions. Experimental results on a 12-month Walmart dataset spanning 15 departments show that the model reduces Mean Absolute Error (MAE) by 18% compared to exponential smoothing baselines, while the spline-enhanced neural component achieves a 24% accuracy improvement over standalone ARIMA. The anomaly detection module successfully identifies 92% of simulated irregularities (e.g., holiday demand spikes) with only a 7% false-positive rate. By uniting statistical rigor with machine learning adaptability, the proposed framework delivers three major advantages: (1) scalable department-level modeling without per-product customization, (2) real-time adaptability to evolving demand patterns, and (3) cost-efficient inventory optimization through integrated anomaly alerts. Overall, this research provides a practical and scalable blueprint for retailers aiming to enhance forecasting precision, mitigate supply chain risks, and reduce operational inefficiencies in dynamic market environments. Keyword: Retail forecasting, hybrid ARIMA–neural network model, real-time demand prediction, anomaly detection, inventory optimization, time series analysis, cubic spline interpolation, residual analysis, Walmart sales data, machine learning, supply chain management.

Similar Items