TEAM
Recipe Recommender
Nadir Hajouji, Felix Almendra Hernandez, Nathan Schley, Ali Arslanhan, Katherine Martin

We designed a recipe recommendation engine that suggests recipes based on a user query and a user's review history.
Our modeling focused mainly on trying to predict recipes that a user was likely to review. We tried some intuitive things, and they didn't work as well as we thought they would- but we obtained models that did a surprisingly good job predicting which reviews were left out of the training set using singular value decomposition.
We also created a user interface that allows a user to enter a freeform query, and that returns a list of recipes that not only match the query but also take the user's review history into account. We did this by combining our model (which quantified how well a recipe matches up with the user history) with a pretrained sentence transformer (which quantified how well a recipe matches the query).


