<![CDATA[Local e-commerce often faces difficulties in providing relevant product recommendations due to diverse catalogs and heterogeneous user preferences. This study designs a hybrid recommendation system for VariasiMotorJKT by combining Market Basket Analysis (MBA) and Collaborative Filtering (CF). The data used are Completed orders from the VariasiMotorJKT store on Shopee for the period July–August 2024 with key attributes such as product name, quantity, buyer username, order time, and total price. The MBA component is extracted using Apriori/FP-Growth to build a “Frequently Bought Together” module, while the CF component (primarily user-based) generates personalized “For You” recommendations. Both are combined through a weighting/reranking scheme and stock filtering so that recommendations are ready to be displayed on product and cart pages. The evaluation is planned offline using chronological data sharing (train/validation/test) and Precision@K, Recall@K, MAP/NDCG@K, and Coverage metrics. The system is expected to improve product navigation, cross-sell/up-sell opportunities, and the quality of store owners’ merchandising decisions. The prototype was developed in a web environment (HTML/CSS, XAMPP) with a simple ETL pipeline for cleaning and basket/interaction matrix generation. The research results are expected to provide practical contributions for e-commerce SMEs and academic contributions as a case study for the application of data mining techniques for recommendations.]]>
