Enhancing the wine tasting experience using greedy clustering wine recommender system

• DOI: 10.1007/s11042-021-11300-5
• Publication Date: 2021-09-17T18:21:13Z
• Authors: Rahul Katarya; Rajat Saini
• Keywords: 0202 electrical engineering, electronic engineering, information engineering; 02 engineering and technology

Wine is not just about taste; it represents your class & overall personality. But tasting the same wine for a long time gets boring; sometimes, everyone needs something new, and for that, we usually get advice from friends. But it is not for sure that their taste will match yours and hence you might not like the wine they like. To resolve this issue, we propose the Greedy Clustering Wine Recommender System (GCWRS). This adaptive wine recommendation system uses principal component analysis (PCA) and K-Means clustering algorithms and a novel ranking algorithm using a greedy technique to recommend wines. Along with that, it also uses the elbow method, which helps find out the ideal number of clusters. Together with the clustering algorithm, the recommendation system is tailored in such a way to help all types of users (the first time tasters or regular wine drinkers) as well as the needs of the users (try something new or enhance their current preferences). The clustering model is trained on the wine dataset (with 6497 entries and 12 attributes) combined with the personal preferences of the user, results in an effective personalized recommendation system. The performance of the model is compared with other standard algorithms like K-means clustering, Agglomerative clustering, DBSCAN, Birch clustering algorithm, using performance evaluation functions like the Silhouette coefficient, Calinski-Harabasz Index, and Davies-Bouldin index. Results indicate that our Greedy Clustering Wine Recommender System (GCWRS) provides a better result than other standard algorithms with Silhouette coefficient=0.28, Calinski-Harabasz Index=2207.84, and Davies-Bouldin index=1.36.