none
User features and Item features do not have effects on Matchbox recommendation

    Question

  • Hi,

    I am trying to train a Matchbox recommendation for less than 1000 users and about 2000 items. The results are bad in general as only a few items are recommended to all the users. Furthermore, I don't see the user features or item features have any effect on the recommendation. I have tried multiple numbers of traits, iterations and training batches.

    Kindly let me know how I can improve the result.

    Thanks

    Monday, January 7, 2019 4:03 AM

All replies

  • Hi,

    Sorry to hear you have problem when using Matchbox recommender, since I don't have the details about your experiment, I can't tell what the problem is. But I would recommend you to refer to our sample experiment: https://gallery.azure.ai/Experiment/3a02931f94114f47b4512dd9179b515e

    And also, if you want to know more details about Matchbox Recommender module, please refer to: https://docs.microsoft.com/en-us/azure/machine-learning/studio-module-reference/score-matchbox-recommender

    Regards,

    Yutong

    Tuesday, January 8, 2019 12:54 AM
    Moderator