A group of 6 UC Berkeley students recently teamed up to tackle an issue they all faced in the kitchen each day: deciding what’s for dinner. With limited ingredients in their typical college apartments, the students dreamed of a platform where they could share and discover new, delicious, yet simple, recipes. Over the course of 3 months, they created a unique Spotify-for-recipes platform like no other from scratch. The platform supports user interactions for sharing recipes with friends and loved ones, fostering a sense of community. Various recipe playlists are individually curated and encourage users to try new foods. This is the perfect platform for all you foodies who are dying to try a fun, new recipe!

Recipe Recommendation Approach

However, all this couldn’t have been accomplished overnight. The soon-to-be grads scraped and cleaned over 4,000 recipes for their database from various websites. Several machine learning techniques such as bag-of-words vectorization and k-means clustering were utilized to create the optimal recommendation model. Recipes are divided into 30-40 labeled clusters that users rate before receiving a recipe recommendation from their top-rated cluster.

The students ultimately succeeded to create this distinct platform because of their shared love for cooking and passion for data science.

Project by: Samuel Huang, Remi Luyssaert, Jason Rudianto, Maria Sooklaris, Tasnim Tallman, Alice Wei