TripAdvisor Recommender System

Inspiration

We wanted to create an efficient system to recommend hotels on TripAdvisor. We wanted to get to the point where we could answer the question, "Given this hotel, which hotels are most similar?" We got there.

What it does

We created a new recommender system based on past user behavior. The new system is able to provide hotel recommendations to users based on the number of viewers that hotels have in common. We answer the question, "What hotels did people who saw this also view?"

How we built it

We primarily used Pandas in Python to run our model (based off of the correlation matrix) and our k-means visualizations (for validation)

Challenges we ran into

With the exception of one of us (who has done Python and ML for half a year) we had pretty much zero previous experience with anything so we had to learn everything by scratch and on our own. And we were all sophomores. Quite the obstacle!

Accomplishments that we're proud of

3/4 of our members did not have Python installed on our computers. We are most proud of how fast we learned, and the things we were able to accomplish in 24 hours.

What we learned

We learned almost everything from scratch. Pandas, Clustering, Python, everything.

What's next for TripAdvisor Recommender System

We created the simplest part of any underlying recommendation model. We could build it up and tweak it to profile users by their history--creating a more robust and complex model depending on what TripAdvisor wants--and we can create a robust frontend that can make the user experience for hotel searches quick and easy. At this point, we need to work with TripAdvisor to figure out where they want our recommender system to go.

Built With

• python