Can the genre of a song be predicted based off of a few audio features? What if more artist-based features are included?
Using a 10,000 song sample, from the Million Song Dataset (https://labrosa.ee.columbia.edu/millionsong/), these questions will attempt to be answered.
This dataset contains a million songs from 1922-2011, with artist tagged information from Echonest (now part of Spotify), along with audio measurements, and other relevant information.
Thierry Bertin-Mahieux, Daniel P.W. Ellis, Brian Whitman, and Paul Lamere. The Million Song Dataset. In Proceedings of the 12th International Society for Music Information Retrieval Conference (ISMIR 2011), 2011.
-The extraction of HDF5 source files and some initial EDA:
01_Extract_HDF5_Files_to_a_CSV_File.ipynb
-Web-scraping the Billboard Top 100 charts from 1958-2011, to determine if a song was popular or not:
02_Scrape_Billboard_top_100.ipynb
-Exploratory Data Analysis:
03_EDA.ipynb
-Predicting if a song is hot/popular or not using information from the Billboard Top 100 charts:
04_Predict_Hot_Billboard Songs.ipynb
-Predicting if a genre can be found with audio features:
05_Genre_Identification.ipynb
An Executive Summary is included in the Documents folder:
Summary
Below is a dictionary of audio terms and columns that are found in the 10k dataset file, for reference:
| Field name | Description |
|---|---|
| artist familiarity | algorithmic estimation |
| artist hotttnesss | algorithmic estimation |
| artist id | Echo Nest ID |
| artist latitude | latitude |
| artist location | location name |
| artist longitude | longitude |
| artist name | artist name |
| artist terms | Echo Nest tags (genre classification) |
| artist terms freq | Echo Nest tags freqs (how often it shows up) |
| artist terms weight | Echo Nest tags weight (genre weight) |
| bars start | beginning of bars, usually on a beat |
| beats start | result of beat tracking |
| duration | in seconds |
| end of fade in | seconds at the beginning of the song |
| key | key the song is in |
| loudness | overall loudness in dB |
| mode | major or minor |
| release | album name |
| sections start | largest grouping in a song, e.g. verse |
| segments loudness max | max dB value |
| segments loudness max time | time of max dB value, i.e. end of attack |
| segments loudness max start | dB value at onset |
| segments pitches | chroma feature, one value per note |
| segments start | musical events, ~ note onsets |
| segments timbre | texture features (MFCC+PCA-like) |
| song hotttnesss | algorithmic estimation |
| start of fade out | time in sec |
| tatums start | smallest rythmic element |
| tempo | estimated tempo in BPM |
| time signature | estimate of number of beats per bar, e.g. 4 |
| title | song title |
| year | song release year from MusicBrainz or 0 |