Setup
Download the source package of MyMediaLite, and open the MyMediaLite solution file src/MyMediaLite.sln
in MonoDevelop.
Hit Ctrl-F8 to build the complete project.
In the MyMediaLite root directory, enter
make download-imdb make download-movielensThis is what you would do on Linux and OS X.
On Windows, you have to perform some steps manually (sorry for that, we intend to change this in the future):
- Download http://www.grouplens.org/system/files/ml-1m.zip and ftp://ftp.fu-berlin.de/pub/misc/movies/database/german-aka-titles.list.gz
- Unzip the two files with whatever archiving program you have (e.g. WinZip). The IMDB stuff should go into data\imdb, the MovieLens stuff into data\ml-1m.
- then run perl scripts/import_dataset.pl --separator=:: data/ml-1m/ratings.dat > data/ml-1m/ratings.txt on the command line
To train the initial model, execute
mkdir data/models rating_prediction --training-file=data/ml-1m/ratings.txt --recommender=BiasedMatrixFactorization --recommender-options="num_iter=60 bias_reg=0.001 regularization=0.045" --save-model=data/models/ml-1m-bmf.modelThen you can run the GUI demo e.g. by opening
src/MovieDemo/Program/Main.cs
and hitting Ctrl-F5 in MonoDevelop.