Cook It!

Cook It! is a web service that is personalized to your tastes and for your taste.

Comment

Cook It!

Cook It! is a web service that is personalized to your tastes and for your taste. It uses the Amazon AWS Machine Learning API to learn your food preferences and to recommend recipes that you can make with the ingredients in your fridge. Just enter your ingredient list and select your meal type (from Breakfast, Main Course, and Dessert), and simply choose your dish from the many recipes that Cook It! has to offer.

Inspiration

Being huge foodies, and very recently overworked college students, making delicacies that could satisfy our palates while being practical at the same time had started becoming increasingly impossible in these last few months. That is when we decided to make Cook It, something that would help us in our food exploration.

How does it Work

We've collected data from two of the largest food recipe sources on the internet, Yummly and Spoonacular and ran Amazon AWS' industry standard regression on it to create an ML model that predicts the correlational success of a given set of ingredients. Moreover, this model evolves over time based on the user's own personal choices and the recipes he chooses to click on. All of this invisible to the user, all one has to do it enter a list of ingredients he might have on hand and wait for the magic to happen. Using web ratings and the past user recorded data, our algorithm creates a sorted list of recipes for the user to choose from starting from the top left.

Challenges we Faced

Being just freshmen, exploring the field of ML was especially hard for us. Applying this to a genre like food where subjectivity prevails and reliable data was extremely hard to find, we had to hand sort a lot of our sources and train our model on around 10,000 existing ingredient combinations and their ratings derived from social networks to achieve a reliably consistent prediction model.

Integrating, consolidating and making the different technologies work together was another aspect that gave us a huge challenge.

Accomplishments that we're Proud Of

Making something that we and our friends are extremely excited to use on a daily basis!

What's Ahead

While our ML model is reasonably reliable right now, we aim to include a few more datasets and run some more training to make it better. We are also planning to improve our recipe generation to get better suggestions.

Share this project:

Updates