Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Added cold-start example along with instructions to build user/item features #591

Open
wants to merge 1 commit into
base: master
Choose a base branch
from

Conversation

V-Sher
Copy link

@V-Sher V-Sher commented Feb 9, 2021

Working example to provide an in-depth explanation of how to create user/item features and demonstrate the use of Lightfm for solving the cold start problem. Also attached link to an article to provide further clarity to LightFM users.

@SimonCW
Copy link
Collaborator

SimonCW commented Feb 20, 2021

Hi @V-Sher ,

thanks for the PR. I think it is a good idea to tackle the topic of building user/item features. This part of the library seems a bit unclear to many users. However, I don't think that it would be a good idea to link to a private medium article (that might even be inaccessible behind the paywall).

Maybe you can try focusing the notebook on the issue of creating user/item features and directly add the necessary explanations in the notebook markdown?

Another way would be to improve upon the official example on hybrid / cold-start to illustrate the process of building user features.

@V-Sher
Copy link
Author

V-Sher commented Feb 20, 2021

@SimonCW Ahh, fair point. I will try to add all the necessary explanations in Markdown in the next couple of days!. Thanks for the suggestion.

@ankurrajdev
Copy link

ankurrajdev commented Aug 24, 2021

What if the user features contain non-binary values such as total orders or avg order value?
We would be passing all unique values for that variable in uf argument.

The current example only has binary input.

Would this approach work for that kind of input for user_features?

Also, just wanted to confirm, if we want to evaluate the recommender. The user_matrix that is passed in the training step should be different from the one passed in predict/evaluate step. Else, it can cause data leakage, if we use variables such as avg order value in last x days.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants