0

Wine Recommender System

TF-IDF + DBSCAN clustering for terroir-based wine recommendations.

Overview

Re-implemented and extended Geronimo Wolf’s research by web scraping wine descriptions and geographic data, then applied TF-IDF and DBSCAN clustering—tuning hyperparameters for sommelier-level recommendation accuracy.

Features

  • Data Enrichment: Scrapes terroir features like soil & climate
  • TF-IDF Analysis: Converts descriptions into feature vectors
  • DBSCAN Clustering: Groups similar wines by geographic & flavor profiles

Built With

  • Python & scikit-learn
  • BeautifulSoup for web scraping
  • pandas & NumPy