Predictive Analytics for Food Retail: Improving Seasonal Sales Forecasts through Advanced Models

Authors

Carlos Martins, Instituto Politécnico do Cávado e do AveFollow
Óscar Ribeiro, Instituto Politécnico do Cávado e do AveFollow
Patrícia Leite, Instituto Politécnico do Cávado e do AveFollow
Paulo Teixeira, Instituto Politécnico do Cávado e do AveFollow
Joaquim P. Silva, Instituto Politécnico do Cávado e do AveFollow

Abstract

Accurate sales forecasting is crucial for success in the retail sector, influencing stock and resource allocation. However, forecasting is challenging due to seasonal patterns. This study enhances forecasting capabilities for a major South American food retailer. After preprocessing and exploratory analysis, several forecasting algorithms were tested, including statistical models, neural networks, and hybrid approaches. Results indicate that LSTM and Prophet-based models significantly outperformed the existing solution, with Prophet recommended for implementation due to cost-effectiveness. The study highlights the importance of seasonal forecasting to improve sales outcomes and reduce food waste.

Recommended Citation

Martins, Carlos; Ribeiro, Óscar; Leite, Patrícia; Teixeira, Paulo; and Silva, Joaquim P., "Predictive Analytics for Food Retail: Improving Seasonal Sales Forecasts through Advanced Models" (2025). CAPSI 2025 Proceedings. 2.
https://aisel.aisnet.org/capsi2025/2