Inspiration

With TripAdvisor's activity and hotel data sets, we set forth to understand and visualize the data using dc.js and python visualization methods. Then we pursued analyzing the data by building a recommendation engine.

What it does

Currently, the recommendation engine and front end aren't connected. The recommendation engine has been trained and can recommend the top three hotels to 1/3 of users from the hotel data set. The front end was set up to show a New York City map (d3.js) to pin hotels and attractions (planned to use Google API).

How I built it

We used python and sklearn for the recommendation engine along with the advice from a TripAdvisor data scientist. We built out the front end using simple javascript and html, which we tested by serving up through a http python server then created a react app for better updates between ui and state.

Challenges I ran into

None of us had extensive experience with machine learning so most of our time was trying to understand what we were attempting to code for the recommendation engine. A lot of us also didn't have experience with React nor d3, but were able to get a New York City map finally with the areas listed in the hotel data set.

Accomplishments that I'm proud of

We learned a lot of new frameworks, concepts, and libraries in this short hackathon. We're really proud we got far in the recommendation engine - creating a base foundation that we can continue to learn from. In addition, we set up a front end. A lot of what we were missing was tying things together (through feeding the data from the recommender to a database like firebase so that it could have access in our web app, which we had more experience with... but ran out of time to implement). We're proud we spent most of our time challenging ourselves with data science concepts and front end data visualizations we weren't familiar with.

What I learned

Scoping our project is important and understanding the data set through visualizations are really helpful! Now we know more about applying machine learning and front end.

What's next for TripViz

Well, now there's a foundation for the entire team to learn from as well as a python recommendation engine and react & d3 front end for anyone else on the internet to learn from!

To try it out, go to our browndatathon2019 repo. You can run npm start in the my-app for the d3 viz. You can run python3 recommender.py to test the code (all it does right now is give you the top 3 recommended for a specific user you input - changed via code).

Built With

Share this project:
×

Updates