BACKGROUND

Billboard announces the Hot 100 chart at the end of each year. This project was based on these charts.

I used Kaggle's file including rank, song name, artist, released year, and lyrics.

Ref :

https://www.billboard.com/charts/year-end

https://www.kaggle.com/rakannimer/billboard-lyrics

MIDI files were used to learn and make music. These files were collected through the internet.

Ref :

https://en.wikipedia.org/wiki/MIDI

https://freemidi.org/

I used the LSTM Recurrent Neural Network (RNN) to generate new melodies. Especially, I used Bannon Dorsey's MIDI-RNN using the Keras library.

Ref :

https://brangerbriz.com/blog/using-machine-learning-to-create-new-melodies

https://github.com/brannondorsey/midi-rnn

Minimal post-production was progressed by humans. Songs were composed only with built-in instruments of Presonus DAW(Digital Audio Workstation) from Studio One.

Ref :

https://en.wikipedia.org/wiki/Digital_audio_workstation

https://www.presonus.com