This project aims to use music listener databases to find new music for you to enjoy.
Currently, I have been using the Last.fm 360K dataset. I plan to use abstractions that allow for other datasets as well.
This project currently supports the following music setups:
- iTunes on Mac OS X
- Spotify on any operating system
If you use a different music player, you may file an issue on this repository and I (or somebody else) may add support for your setup.
First, download the Last.fm music dataset. While this is running, make sure you have Go installed and a GOPATH configured.
To download the tool, run:
$ go get github.com/unixpickle/goodmusic/recommender
Now, follow the instructions on iTunes or Spotify below.
Run the command as follows, and you will see similar output:
$ $GOPATH/bin/recommender itunes /path/to/usersha1-artmbid-artname-plays.tsv 100 | more
2016/09/22 19:35:40 Loading music library...
2016/09/22 19:35:41 Loading dataset...
2016/09/22 19:36:06 Building music profile...
2016/09/22 19:36:06 Computing artist correlations...
2016/09/22 19:36:09 Sorting correlations...
2016/09/22 19:36:09 Recommending...
the killers (correlation of 77279.27%)
the beatles (correlation of 65038.06%)
britney spears (correlation of 63985.37%)
red hot chili peppers (correlation of 63947.08%)
...
The last argument (100 in this case) specifies how many plays an artist needs in iTunes in order to be considered. I have found that the default of 100 works fine, since most of my favorite artists have hundreds of plays and I have few artists in iTunes that I don't enjoy.
Run the command as follows, and you will see similar output:
$ go run *.go spotify ~/Downloads/lastfm-dataset-360K/usersha1-artmbid-artname-plays.tsv 1
2016/09/23 19:12:21 Loading music library...
Please navigate to:
https://accounts.spotify.com/authorize?client_id=dac4cf3243db4bd1b055efc7af84be72&redirect_uri=http%3A%2F%2Flocalhost%3A14505%2Fspotifydone&response_type=token&scope=user-library-read
2016/09/23 19:12:28 Loading dataset...
2016/09/23 19:13:06 Building music profile...
2016/09/23 19:13:06 Computing artist correlations...
2016/09/23 19:13:09 Sorting correlations...
2016/09/23 19:13:09 Recommending...
the killers (correlation of 77279.27%)
the beatles (correlation of 65038.06%)
britney spears (correlation of 63985.37%)
red hot chili peppers (correlation of 63947.08%)
...
The last argument (1 in this case) specifies how many songs you must have by a given artist for that artist to be considered.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent