Recommender API for skinsearch.co, created in Summer 2021 for MIDS Capstone Other collaborators: @dineshachuthan @xtinachen
My main contribution were the .py files in the utils folder. These generate recommendations through the flask app in app.py.
Instructions:
- from the RecommenderAPI folder, run
python utils/recommender2.py- this takes the raw review data and does the SVD decomposition/recomposition
- it also does the bag of words on the products data
- it takes cosine similarity within each of the two matrices described above and places the resulting matrix in the data directory
- also from the RecommenderAPI folder, run
python app.pyto activate the Flask app - add to the end of the Flask url any of the following to see the returned recommendation
/WithIngredients/genus/Moisturizers/skin/combination?n_recs=5/WithoutIngredients/genus/Moisturizers/skin/combination?n_recs=5/RoutineWithIngredients/genus/Moisturizers/skin/sensitive- each returns a list of SKU numbers corresponding to the products data in the data directory
This app was called by other components of SkinSearch.