Perfume Match: a new way to find your next perfume.
Perfume Match is a Flask-based descriptive perfume recommendation web-app that I developed during Insight. This folder include all the data and codes.
Please check: www.perfumematchcc.com
Main metrics that were used:
- Scraped, cleaned and structured perfume data including 260 products and total of 88,005 custom reviews
- Performed NLP analysis of perfume description and reviews through doc2vec and TF-IDF word embedding;
- applied singular value decomposition (SVD) for dimension reduction
- Devised new similarity-based recommendation algorithm; recommended perfumes for individual users