Predictive Modeling for Restaurant Menu Customization: An FP-Growth Algorithm-Based Solution

Customizing the eating experience through personalized food recommendations has become crucial in the dynamic restaurant industry. Conventional recommendation systems often struggle to meet individual preferences and ordering patterns. The problem of analyzing and processing the huge amount of historical transactional data to offer tailored food recommendations that align with individual tastes and preferences is still a challenging task. This paper proposes an innovative solution to forge a resilient Food Recommendation Engine framework that would be able to make prediction as a set of five items as recommendation for each sold item of the restaurant. The proposed framework leverages the FP-Growth algorithm as a base algorithm to generate the frequent itemset and association rules for both item and its item category. Over the base algorithm was systematically evaluated across distinct meal categories with different support and confidence threshold and we combined a business-driven filtering rules and a custom prediction algorithm that has capabilities to make prediction on the sold item as recommendation based on the associations rules and frequent itemset obtained for both item and its category from the Fp-Growth. The prediction algorithm also provides some placeholder for adding items as advertisements to facilitate promotions. The generated food recommendation contains the combination of foods and beverages which results the total combination of five items per sold item for each part of the day meal including some placeholders. This research covers the complete framework development process, including data preparation, fp-Growth algorithm implementation, business-driven filtering rules and prediction algorithm design to generate recommendation item lists for each mealtime.