🎧 Song Popularity Prediction 🎹

The digital music industry is evolving rapidly, and streaming platforms compete to estimate how well a song will perform among listeners. Popularity is more than just a number—it influences recommendations, automated playlists, and even artists' contracts.

To train a new prediction module, you have received a dataset extracted from Spotify containing audio features and metadata of songs.

Your goal is to build a model that predicts a song’s popularity based on its numerical characteristics and several auxiliary features.

📁 Dataset

You have two files:

• train.csv — songs with all columns including popularity
• test.csv — songs without the popularity column (these must be predicted)

Each song has a unique identifier and a set of features:

• track_id — unique song ID
• artists
• album_name
• track_name
• popularity (train only)
• duration_ms
• explicit
• danceability
• energy
• key
• loudness
• mode
• speechiness
• acousticness
• instrumentalness
• liveness
• valence
• tempo
• time_signature
• track_genre

The main goal is to predict the popularity column for the test set.

🧠 Task (100 points)

Build a machine learning model that predicts the popularity of each song in test.csv.

Your predictions must be saved in a submission.csv file with the format:

track_id,popularity
101,42.7
102,55.1
103,38.0

where:

• track_id — the song ID from the test set
• popularity — a real-valued (float) predicted popularity score

📊 Evaluation

Evaluation will be performed using MAE (Mean Absolute Error):

The final score is based on the obtained MAE using the following rules:

• MAE ≤ 5 → 100 points
• MAE ≥ 20 → 0 points
• Intermediate values receive proportional scores between 0 and 100.