Machine Learning and Deep Learning for demand forecasting in Logistics: A systematic literature review

In recent years, the increasing complexity of supply chain operations has driven a growing interest in the use of artificial intelligence, particularly deep learning, for demand forecasting. This study presents a systematic literature review (SLR) that identifies, classifies, and analyzes the most commonly used deep learning models in logistics demand planning. Using the PICO methodology, research questions were structured to guide the selection and evaluation of relevant studies. A total of 606 initial sources were retrieved from SCOPUS, and after applying strict inclusion and exclusion criteria (including year of publication, language, access type, publication status, and journal quartile), 86 articles were included in the final analysis. The results reveal significant academic interest in the topic, with a peak in publications in 2023. Models such as CNN, LSTM, GRU, and hybrid approaches demonstrate high forecasting accuracy and adaptability to complex logistics environments. However, its performance depends on factors such as data quality, operational context, and model configuration. The review highlights current trends, identifies research gaps, and suggests future courses of action focused on improving data integration and developing context-specific hybrid models. This analysis provides a solid foundation for researchers and practitioners seeking to implement deep learning techniques in logistics and supply chain demand planning.