Integration Of Customer Preferences And Ordering Patterns For Menu Recommendation Personalization Using Topsis

Abstract

Micro, Small, and Medium Enterprises (MSMEs) in the Indonesian culinary sector frequently face operational bottlenecks and customer decision fatigue during peak dining hours due to manual menu selection and conventional order-taking mechanisms. Preliminary field observations at Pondok Orange Restaurant in Pontianak revealed that customer menu deliberation required 6.5 to 8.2 minutes per table, while ordering queues extended up to 25 minutes during peak meal intervals. To address these operational challenges, this study designs and implements a web-based food menu recommender system utilizing the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS). The recommendation framework integrates two complementary data dimensions: Customer Preferences (incorporating menu selling price as a cost attribute and customer satisfaction ratings as a benefit attribute) and Operational Ordering Patterns (incorporating daily order frequency and real-time kitchen ingredient availability as benefit attributes). Developed using the Extreme Programming (XP) methodology on the Laravel framework, the system executes TOPSIS matrix computations with linear algorithmic complexity O(m × n) in under 5 ms without requiring extensive historical customer logs. Computational evaluation shows that Bubur Pedas achieved the highest preference score of 0.6498, securing the top recommendation ranking. Initial user acceptance testing with 30 dine-in patrons demonstrated strong alignment, yielding a Top-1 recommendation hit rate of 73.3%, a Top-3 hit rate of 90.0%, and an overall usability and relevance satisfaction score of 4.38 out of 5.00. Software verification via White-Box Testing (Statement Testing and Basis Path Testing, V(G) = 4) confirmed that all four independent algorithmic paths and mathematical boundary conditions executed successfully according to technical design specifications.