Customer churn is a critical issue for modern businesses, particularly in highly competitive sectors such as gastronomy. High churn rates, ranging from 30% to 50% in hospitality and food services, highlight the urgent need for strategies that enhance customer retention. This study conducts a Systematic Literature Review (SLR) to identify and analyze machine learning-based models used to predict customer churn, with a focus on their applicability to the gastronomic sector. Predictive models trained on historical consumption, visit frequency, satisfaction levels, and interactions can classify customers by risk, enabling timely interventions. The methodology followed a structured five-step approach, incorporating the PICO framework and PRISMA 2020 guidelines to ensure transparency and reproducibility. The search strategy used a comprehensive Boolean equation across Scopus, retrieving 777 initial articles. After applying defined inclusion and exclusion criteria—such as open-access availability, language, and publication recency—102 studies were selected. The process was documented using a PRISMA flowchart. Key findings emphasize the growing role of machine learning in customer behavior prediction and the importance of methodological rigor in literature synthesis. This review contributes to evidence-based decision-making and offers insights into selecting high-accuracy predictive models tailored to the specific needs of enterprises.