This work have been made in group. Thanks to my crewmates Jacky and Kevin.
The dataset contains features of songs for a certain user collected on Spotify. The goal of this project was to predict if the user will like a track or not based on what he usually likes. We have been ranked based on our model accuracy between others students on Kaggle.
All songs are labeled in the column target: “1” means the user likes it while “0” means he does not.
There is 16 columns in the dataset : The three latest describe the song with its name, the author and if it has been liked. The others are the features with which we will construct our model:
- acousticness.
- danceability.
- duration_ms.
- energy.
- instrumentalness.
- key.
- liveness.
- loudness.
- mode.
- speechiness.
- tempo.
- time_signature.
- valence.
Here is the result of the competition : we finished 1st in a tie with an accuracy of 82% (IJK Team).