Given a playlist curated from the MillionSongDataset, give song recommendations that are similar to the playlist. A subset of 10,000 random songs were taken from the dataset to be used in our analysis.
Create a virtual environment and cd in
python3 -m venv <name_of_venv>
cd <name_of_venv>
Activate the venv which changes your paths
source ./bin/activate
Clone the repository and cd in
git clone git@csil-git1.cs.surrey.sfu.ca:bvtrinh/353-project.git
cd 353-project
Install requirements
pip3 install -r requirements.txt
Afterwards run the Jupyter Notebook
jupyter-notebook
Open the index.ipynb and stats.ipynb for song recommendations!
Download the subset from the MillionSongDataset
There will be two folders contained in the main folder:
AdditionalFiles/ data/
Using the scripts in parser/, generate the data files for the system
Aggregate the h5 files to generate from_h5.csv:
python3 h5_to_csv.py data from_h5.csv
Aggregate the db files contained in AdditionalFiles/ to generate from_sql.csv:
python3 db_to_csv.py subset_artist_term.db subset_artist_similarity.db subset_track_metadata.db
Combine the two output files to generate songs.csv:
python3 combine.py from_sql.csv from_h5.csv
Create artist.csv with:
python3 artist_filter.py from_sql.csv
Overwrite the files in data/ with the newly created artists.csv and songs.csv