Banca de QUALIFICAÇÃO: SERGIO GLEDSON DE LIMA MARQUES
Uma banca de QUALIFICAÇÃO de MESTRADO foi cadastrada pelo programa.STUDENT : SERGIO GLEDSON DE LIMA MARQUES
DATE: 12/09/2025
TIME: 15:00
LOCAL: https://meet.google.com/yea-wvfd-gkp
TITLE:
HYBRID SARIMA-AUTOFORMER MODELING FOR RETAIL DEMAND
FORECASTING
KEY WORDS:
Demand forecasting, Time series, Hybrid models, Autoformer, Retail.
PAGES: 73
BIG AREA: Ciências Sociais Aplicadas
AREA: Administração
SUBÁREA: Administração de Empresas
SPECIALTY: Administração Financeira
SUMMARY:
Demand forecasting plays a strategic role in retail management, contributing to inventory
optimization, reduction of operational costs, and improvement of the consumer experience.
This study proposes the evaluation of hybrid models for sales forecasting, with an emphasis
on modeling the residuals generated by traditional models. The research begins with a
literature review, which highlights the predominance of architectures based on Long Short-
Term Memory (LSTM) and Random Forest (RF) as hybridization strategies. However, a gap
is identified regarding the application of the Autoformer model in conjunction with Zhang’s
(2003) methodology. The Autoformer, by incorporating time series decomposition
mechanisms and autocorrelation, shows potential to capture seasonal patterns and long-term
trends. The Seasonal AutoRegressive Integrated Moving Average (SARIMA) is adopted as
the base model, and the Autoformer as the hybridization strategy for modeling the residuals,
whose performance will be compared with recurrent approaches in the literature, such as
LSTM and RF. Using sales time series from the Ecuadorian retail chain La Favorita, the study
evaluates the performance of the approaches through performance metrics. The results are
expected to contribute to methodological advancements in univariate and seasonal time series
forecasting, as well as provide practical applications in demand forecasting.
COMMITTEE MEMBERS:
Interno - DIEGO DE SOUSA GUERRA
Externo ao Programa - 3431335 - JAIR PAULINO DE SALES - nullExterno à Instituição - LUIS ABEL DA SILVA FILHO - URCA
Presidente - PAULO RENATO ALVES FIRMINO