Comparative Analysis of Modern Machine Learning Models for Retail Sales Forecasting
Croatian Operational Research Review (CRORR)
Abstract
Accurate demand forecasting is critical for brick-and-mortar retailers to optimize inventory management and minimize costs. This study evaluates statistical baselines, tree-based ensembles (XGBoost and LightGBM), and deep learning architectures (N-BEATS, N-HiTS, and the Temporal Fusion Transformer) on retail sales data characterized by intermittent demand, substantial missingness, and frequent product turnover. Models are compared across four configurations varying by aggregation level and imputation strategy, using evaluation protocols that reflect typical deployment patterns for each model class. Localized tree-based methods achieve superior performance, with XGBoost attaining the lowest RMSE of 4.833. While SAITS-based imputation improved neural network performance in aggregated settings, these models remained inferior to ensemble methods. The results suggest that, under the studied constraints, model selection should prioritize alignment with problem characteristics over architectural sophistication.
Notes
Published in Croatian Operational Research Review 17(2), 2026 (doi:10.17535/crorr.2026.0021). A comparative study of statistical, tree-ensemble, and deep-learning forecasters on real retail sales data — finding that localized tree-based methods (XGBoost) outperform more sophisticated neural architectures under intermittent demand and heavy missingness.
How to cite
@article{brcic2026brcic,
author = {Luka Hobor and Mario Brcic and Lidija Polutnik and Ante Kapetanovic},
title = {Comparative Analysis of Modern Machine Learning Models for Retail Sales Forecasting},
journal = {Croatian Operational Research Review (CRORR)},
year = {2026},
doi = {10.17535/crorr.2026.0021},
}