Framing Question: Can recipes be recommended given a set of ingredients?
Purpose and Need:
In the UK, 6.7 million tonnes of food is wasted per year (Source). Some of that food wastage can be attributed to ingredients found in our fridges that are left unused for too long. One possible solution to this problem would be to maximise the usage of food in our fridges. This can be done by cooking meals that incorportate such foods. However thinking of good recipes for these ingredients may be difficult. Thus it is valuable to explore whether recipes can be recommended given a set of ingredients.
Alternatively, such a model could be used to find recipes similar to those that one may already like (or contain similar ingredients).
Data Description:
The Recipe1M+ dataset, by MIT will be used for this project.
The dataset includes information on recipes including instructions, ingredients, nutritional information, and name of recipe.
The dataset has 1 million recipes.
Tools:
- Python, pandas, NumPy, SciPy, scikit-learn
- Maybe need cloud computing such as AWS
- Matplotlib, Seaborn, Tableau
- SpaCy
MVP Goal:
Document-term matrix for dataset. Including text preprocessing.
+Topic Modelling
+Recommendation System