gorse icon indicating copy to clipboard operation
gorse copied to clipboard

Issue: Repetitive Recommendations

Open Shubham-Raj-25 opened this issue 1 year ago • 3 comments

Description:

I am currently using Gorse for recommendation in my application, and I've noticed that all users are consistently receiving popular games as recommendations, resulting in repetitive suggestions. This behaviour is unexpected, as I expect the recommendations to be more personalized based on users' interactions.

Steps to Reproduce:

Set up Gorse with the following configuration:

[recommend]
cache_expire = "10s"
cache_size = 100

[recommend.online]
fallback_recommend = ["item_based", "popular"]
num_feedback_fallback_item_based = 4

[recommend.offline]
enable_latest_recommend = true
enable_popular_recommend = true
enable_user_based_recommend = true
enable_item_based_recommend = true
enable_collaborative_recommend = true
enable_click_through_prediction = true
explore_recommend = {"latest": 0, "popular": 0}
refresh_recommend_period = "1m"
check_recommend_period = "1m"

Have multiple users (e.g., 600 users) interact with around 80 games over time. Expected Behavior: I would expect the recommendation system to provide more diverse and personalized recommendations based on users' interactions and preferences.

Actual Behavior: All users seem to be receiving the same popular games as recommendations, leading to repetitive suggestions for everyone.

Note: I have tried various configurations and explored the documentation, but I couldn't find a solution to this problem.

Shubham-Raj-25 avatar Jul 28 '23 09:07 Shubham-Raj-25

Hi @Shubham-Raj-25 , where you able to fix that? how did you fix it please? I'm having same issue #802

Xoshbin avatar Feb 19 '24 20:02 Xoshbin

I think after some time, as the model got enough data, we had personalised recommendations. Initially because of lack of data of user, it gave popular items only. I am not associated with that work anymore thus I don't remember much about it. :(

Shubham-Raj-25 avatar Feb 22 '24 06:02 Shubham-Raj-25

I think after some time, as the model got enough data, we had personalised recommendations. Initially because of lack of data of user, it gave popular items only. I am not associated with that work anymore thus I don't remember much about it. :(

Thank you for replying

Xoshbin avatar Feb 22 '24 20:02 Xoshbin