Inspiration
I was inspired by the story of Databricks being founded during the workshop, along with my desire for better media recommendations.
What it does
Used Databricks to create a media recommendation app for books, movies, shows, music, and videogames. Data is sourced from various Kaggle databases (from Netflix, Spotify, Steam, Goodreads). Users can search the title of any media in the database and see top recommendations and their categories.
How we built it
I found the relevant data sets from Kaggle, sourced from Netflix, Spotify, Steam, and Goodreads. I used Databricks to set up the database with that information and make an algorithm for categorizing/tagging the media. I then used those tags to compare media to find the best matches. Then I set up the UI and functionality for actually searching a piece of media and getting the top 10 matches for each media type, and deployed the app.
Challenges we ran into
I was very limited by the datasets that I found, as many entries were missing some data that would help with categorization. A large issue was standardizing genre categories between very different medias, and forced me to make decisions about the categories that I would most likely want to change later. Some categories don't give many cross-category or cross-media results, which is an issue.
What's next for Databricks Challenge
Major adjustments to the dataset and categories is needed. Many searches still give sparse matches, and the datasets should be increased for more accurate matches.
Built With
- databricks
