Sales forecasting with machine learning: a hybrid approach for the dynamic fashion sector

• Ano de defesa: 2023
• Autor(a) principal: Januário, Leandro Frigo
• Orientador(a): Lourenço, Carlos Eduardo
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Palavras-chave em Inglês: Sales forecasting; Machine learning; XGBoost
• Link de acesso: https://hdl.handle.net/10438/34873

Small and medium-sized enterprises (SMEs) face the significant challenge of accurately forecasting sales to optimize inventory management and maintain brand value. This complexity is exacerbated in the dynamic fashion sector owing to short product life cycles and rapidly changing consumer preferences. This study focuses on comparing and refining sales forecasting methods with traditional methods (Moving Averages), machine learning models (XGBoost), and Integrated Forecasting models. The analysis revealed that while Moving Averages are effective in managing percentage errors and show strengths during stable sales conditions, XGBoost reduces total and absolute quantity errors. The IF model synergistically combines these methods, which frequently surpasses their individual performances. The results emphasize the potential benefits of integrating traditional and advanced techniques for more robust and accurate sales predictions.