Given the completion of building the model, there is still room to improve the project, which can be mainly categorized in two direcitons mentioned below:
- An Interactive User Interface: For future work, we can build an interactive interface so that users can create playlists, and the recommended songs will appear automatically on the interface. There will be buttons of moods on the interface as well. When users click through each mood button, the recommended songs predicted to be in this mood will pop up.
- A Mixed Model: We can also combine our models together so that users can click through a mood button and add several songs to their playlists, and then some new songs will be recommended based on the new playlists. In other words, we would be able to make predictions on a dynamically changing playlist with recommended songs added in.