I am really impressed by the library you have build. I am currently using it with my own dataset to recommend items to users. I do notice however, and this seems to be an issue with most algorithms out there, that the performance of a model is heavily dependent on the side-features used. Some configurations of features seem to work really well while others seem to decimate the performance.
I was wondering if there is any way to retrieve the importance of features using a trained model. This could be helpful in selecting which features should be included in the training.
Is there any model in your library capable of doing this?
I am really impressed by the library you have build. I am currently using it with my own dataset to recommend items to users. I do notice however, and this seems to be an issue with most algorithms out there, that the performance of a model is heavily dependent on the side-features used. Some configurations of features seem to work really well while others seem to decimate the performance.
I was wondering if there is any way to retrieve the importance of features using a trained model. This could be helpful in selecting which features should be included in the training.
Is there any model in your library capable of doing this?