The focus of the project was to explore whether recipes could be recommended given a set of ingredients.
In order to produce a recommendation system, the following steps have been completed so far:
- Obtaining data on recipes
- Formulating the preprocessing stage for recipe ingredients
- Transforming the recipe ingredients through the TFIDF Vectorizer
- Transforming the document-term matrix through the NMF Model
Here are the top 10 ingredients that have appeared in 1 million recipes (counts included):
Here are the top 10 topics and their top 5 terms (NMF):
Future Steps Include:
- Making a recommendation system using the above topics
- Trying SVD
- Trying clustering methods
- Trying the inclusion of recipe instructions