<![CDATA[Fast and accurate product selection remains a primary challenge for retail businesses such as Yusra Elektronik Store, where the recommendation process to consumers still relies on the manual memory of the management. This condition leads to inefficient service, especially given the large variety of available electronic products. Therefore, this study implements a product recommendation system based on Content-Based Filtering to automatically present relevant alternative items based on their specifications. The system approach is built using the Term Frequency-Inverse Document Frequency (TF-IDF) method for weighting the extracted product description text, along with the Cosine Similarity algorithm to calculate the vector angle distance between products. The dataset processed in this study consists of 217 products divided into 18 main categories of the store’s inventory. The algorithmic computation is executed via a REST API using the Python programming language, while the user interface is developed using PHP. The test results prove that the system successfully presents recommendations independently without requiring past user transaction history data, thereby effectively overcoming the cold-start problem. The algorithm’s performance evaluation based on sample queries recorded an average precision rate (Precision@3) of 93.3% and an average recommendation coverage rate (Recall@3) of 64.5%. Furthermore, the expert judgment validation from the store owner confirms that the system’s recommendation output aligns with the alternative product offering patterns typically conducted face-to-face, and the developed interface is considered highly helpful in accelerating inventory management operations.]]>
