Inventory optimization for Pingo Doce & Go Nova - a deep reinforcement learning approach

• Autor(a) principal: Belchior, Rodrigo Castelo Branco
• Data de Publicação: 2024
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositórios Científicos de Acesso Aberto de Portugal (RCAAP)
• Texto Completo: http://hdl.handle.net/10362/174968
• topic: Drl = deep reinforcement learning; Ppo = proximal policy optimization; Mdp = markov decision process; Ml = machine learning; Pd&G = Pingo Doce & Go; Dqn = deep q- learning; Rl = reinforcement learning; Domínio/Área Científica::Ciências Sociais::Economia e Gestão

Addressing PD&G’s inventory management inefficiencies, this research evaluates the in tegration of predictive analytics and deep reinforcement learning (DRL) to navigate the store’s unique demand influenced by the academic environment. Traditional MRP sys tems’ shortcomings are addressed by implementing SARIMAX, XGBoost, and Neural Prophet models for demand forecasting, alongside DQN and PPO for stock replenish ment optimization. Results demonstrate that advanced forecasting models and DRL may greatly enhance the accuracy of inventory management in comparison to the currently used practices. The deployment of these sophisticated models not only enhances PD&G’s operational efficiency but also pioneers innovative practices in retail inventory management.