<![CDATA[The growth of e-commerce in the fashion sector has driven the need for systems that can help users discover products more efficiently amid an abundance of available choices. This study aims to implement a smart recommendation system based on content-based filtering on the Dinara Konveksi e-commerce website using the TF-IDF (Term Frequency–Inverse Document Frequency) method and cosine similarity.The dataset consists of 35 active products registered in the Dinara Konveksi e-commerce system. The data used consists of product attributes including name, category, description, variant colors, and sizes, which are combined into text documents and processed through preprocessing steps comprising case folding, tokenization, and stopword removal. Feature weighting is performed using TF-IDF to generate a vector representation for each product, while cosine similarity is used to measure the degree of similarity between products. The system produces two types of recommendations: similar product recommendations displayed on product detail pages, and personalized recommendations tailored to users' purchase history, shopping cart, and product view history. The implementation uses a two-tier architecture consisting of a Python script with scikit-learn for batch computation and a PHP service as the runtime interface. Evaluation results show a Precision@4 of 0.75, Recall@4 of 0.60, F1-Score of 0.67, and a Hit Rate of 1.00. The developed system is capable of providing relevant and responsive product recommendations while supporting a more personalized shopping experience for users of the Dinara Konveksi website.]]>
Copyrights © 2026
