<![CDATA[Indonesia's remarkable diversity of coffee sensory profiles presents both an opportunity and a challenge: consumers often struggle to identify coffees that match their taste preferences due to the absence of a recommendation system that aligns choices with measurable sensory attributes. This study develops an Indonesian coffee recommendation system based on K-Means Clustering using a dataset of 50 Indonesian coffee sensory profiles automatically collected through web scraping from the professional coffee review platform Coffee Review. The dataset comprises five quantitative sensory attributes: Aroma, Acidity, Body, Flavor, and Aftertaste. After applying Min-Max normalization, the K-Means algorithm with k-means++ initialization identified three optimal clusters (K = 3), determined using a combination of the Elbow Method, Silhouette Score (0.4889), Davies–Bouldin Index (0.831), and Calinski–Harabasz Score (45.41). The system recommends the top three coffee products by calculating the Euclidean distance between user preferences and the centroid of the nearest cluster. Functional evaluation using Black Box Testing across five test scenarios achieved a 100% success rate (5/5). This study contributes an objective sensory profile–based recommendation approach that addresses the cold-start problem in coffee recommendation systems and demonstrates strong potential for integration into local coffee applications and Indonesian coffee e-commerce platforms.]]>
Copyrights © 2026
