Inspiration

As a college student, it can be difficult to find easy and cheap recipes that are also unique and delicious. Studies show that many college students struggle with finding and affording healthy food. This affects as much as 46 percent of community college students and 40 percent of students attending a four-year university. Additionally, many dining halls on college campuses may not always have the healthiest or freshest options or cater to one’s likes and/or food allergies.

What it does

Our solution is a mobile app called Hungr which can be used as a meal plan substitute and allow students to find their own budget-friendly and healthy recipes for foods with fresh ingredients. Our app takes in the user’s preference for ingredient and type of cuisine and uses a machine learning model to predict affordable recipes from a database. The user can select recipes for the week that they wish to make based on suggestions. Once their desired number of recipes are selected, a shopping list with an estimated price is given for quick and easy access when grocery shopping for the week.

How we built it

Our project was built using 3 different parts. The backend consists of a web scraper that is written in node.js. It uses Puppeteer to navigate through different websites to find and collect the best recipes. It then stores them in a standardized format that can be used later on. The server then accepts the preferences of the user and uses them to train a machine learning decision tree regressor model to predict new recipes that the user may like. We decided to use an explainable AI like decision trees rather than something more complex like neural networks. At the start of every week, the client presents the user with the previous week’s recipes and prompts the user for feedback. It passes this feedback to the server and presents new recipes reflecting the user’s preferences in the form of an iOS app.

Challenges we ran into

Initially, we worked on the frontend and backend separately so it was challenging to coordinate the two when we got to that point. Additionally, the machine learning model and Python aspects of our project required us to tune and train the model quite a bit before getting the most accurate predictions for our users based on their preferences.

Accomplishments that we're proud of

We are happy with how seamlessly we were able to incorporate the information we gathered when web scraping to the machine learning model and ultimately to the frontend for the user to view.

What we learned

For most of our group members, this was the first time that they had used Swift to make a mobile app. This involved quite a bit of design thinking and UI layout planning which is a different way of thinking than just writing the code for a project like this.

What's next for Hungr

In the future, we plan to continue to tune our model to get the most accurate results possible in our meal predictions. We also would like to add new features such as automatic shopping cart redirect to sites like Amazon or HEB that allow the user to purchase their grocery list online rather than in person.

Share this project:

Updates