Spotify Feature Learning

Details

Youtube Demo Link
Group Members: Brandon Fujii, Casey Poon, Grant Yu, Jonathan Huang, Devin Bajaj, and Crystal Gong

Data

• training dataset
• testing dataset
• complete dataset

File descriptions

• \archive contains:
  • Cross Entropy.ipynb contains three attempts at cross entropy classification & ugly feature histograms
  • spotify_genre_classifier.ipynb contains first KNeighborsClassifier attempt
• 2_genre_classification_classical_and_rock.ipynb classifies classical music and rock music based on energy and danceability
• 2_genre_classification_jazz_and_country.ipynb classifies classical music and rock music based on acousticness and loudness
• knn.ipynb contains KNeighbors classification trial
• linear_regression.ipynb features scatter plots comparing each pair of features and shows relationships between: loudness and energy, acousticness and Loudness
• common.py contains commonly used functions (gradient descent, models, plot cost history)
• knn.ipynb contains a knn attempt
• song_feature_distributions.ipynb features prettified feature histograms for all genres and by genre
• 2_means_clustering.ipynb used k-means clustering for classification of two genres (classical and rock) over 2 attributes (energy and danceability)
• 2_means_clustering_3_dim.ipynb used k-means clustering for classification of two genres (classical and rock) over 3 attributes (energy and danceability and acousticness)

Refer here for details on how to retrieve Spotify feature data from other playlists.