McDonalds and Culture

Similar countries in cluster by Culture
Cultural Profile
McDonalds menu recommendations
Dashboard View

Inspiration

McDonalds adapts their menu in different countries to accommodate cultures. Instead of offering the same menu items across the globe, we wondered if there was a way to recommend menu items using culture data. For countries that don't have McDonald's, we wanted to predict what items would appear on a McDonald's menu if it were to open there.

What it does

Users can select countries that don't have McDonald's locations and see the cultural profile, the countries that are close to them culturally, and recommended McDonald's menu items using our ML model. Users can also see the most relevant features that are used in making recommendations. Users also have the option to change the country parameters to their liking and see the recommendations the model makes.

How we built it

The Cultural Menu Optimizer takes in a country's cultural profile like religion composition, meat consumption habits, and economic indicators and runs it through a two-stage pipeline. First, a Random Forest classifier predicts which menu cluster the country falls into (Beef Dominant, Chicken Dominant, Vegetarian Heavy, Seafood Adapted, or Balanced Western). Then, K-Nearest Neighbors finds the most culturally similar countries that already have McDonald's, and we aggregate their menus weighted by similarity score before returning the final recommendations with confidence scores.

Challenges we ran into

Finding the right dataset was a challenge. We also initially had trouble figuring out how to train the model, since using culture as a general feature to predict McDonald's menu items isn't straightforward. To solve this, we broke culture down into specific measurable aspects like religion, dietary habits, and economic indicators and quantified them as our model inputs.

Accomplishments that we're proud of

We've produced some pretty interesting visualizations with this project. We're proud of the fact that you can filter by country, and the results give an insight into their cultures through the lens of McDonalds items. We are also happy to have included a number of advanced machine learning methods, and like how they translated into the final project.

What we learned

We learned how to take something as broad as culture and encode it into specific, measurable features that a machine learning model can actually work with. We also got a better understanding of how K-Nearest Neighbors works as a recommendation engine, using cultural similarity to find neighboring countries and letting their existing menus inform what a new country should get.