IMPLEMENTATION OF THE APRIORI ALGORITHM FOR COFFEE DRINK PRODUCT RECOMMENDATIONS WITH PAYMENTS USING DYNAMIC QRIS

Abstract

Most small coffee shops still record transactions manually, which is error-prone, slow during peak hours, and produces no sales reports. Static QRIS payments require customers to type the amount themselves, risking input errors, while commercial cashier applications generally charge monthly fees. This study aims to design and build a mobile-friendly web-based cashier application supporting cash and dynamic QRIS payments. The application is built with Next.js 14 and a PostgreSQL database accessed through Prisma ORM. The QRIS feature is implemented over two channels: converting the merchant's static QRIS into dynamic QRIS according to the EMVCo standard, and integrating the Midtrans payment gateway with automatic payment status updates via webhooks. The application also provides self-ordering through table QR codes and server-enforced out-of-stock marking. The Apriori algorithm is applied to transaction history to mark best-selling items in each category and to derive association rules for upselling recommendations. Black box testing of 23 scenarios shows that all test cases behave as expected. The results show that the application speeds up transaction recording, shortens queues, eliminates QRIS amount-typing errors, and automatically provides daily sales reports.