{
  "id": 363925,
  "title": "The relatively less data for each test session",
  "url": "/competitions/otto-recommender-system/discussion/363925",
  "author_name": "",
  "post_date": "2022-11-03T17:44:59.079009800Z",
  "votes": 5,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I find that the amount of data(actions) in each test session is relatively less(with a large proportion containing only 1 or 2 actions) compared to that in a training session, so it's difficult to identify the feature of the user itself. Does it mean we can hardly generate a label for a certain type of user,  but should focus more on generating labels for the goods/aids?</p>",
  "messages": [
    {
      "id": "2016034",
      "postDate": "11/03/2022 17:44:59",
      "content": "<p>I find that the amount of data(actions) in each test session is relatively less(with a large proportion containing only 1 or 2 actions) compared to that in a training session, so it's difficult to identify the feature of the user itself. Does it mean we can hardly generate a label for a certain type of user,  but should focus more on generating labels for the goods/aids?</p>",
      "rawMarkdown": "I find that the amount of data(actions) in each test session is relatively less(with a large proportion containing only 1 or 2 actions) compared to that in a training session, so it's difficult to identify the feature of the user itself. Does it mean we can hardly generate a label for a certain type of user,  but should focus more on generating labels for the goods/aids?",
      "votes": null
    },
    {
      "id": "2016236",
      "postDate": "11/03/2022 21:11:27",
      "content": "<p>The reason for the shorter sessions is explained <a href=\"https://www.kaggle.com/competitions/otto-recommender-system/discussion/363554#2015486\" target=\"_blank\">here</a> by the organizer.</p>\n<p>I don't think you can do too much about inferring things about the user as there are no user ids! But maybe you can come up with some smart clustering approach to the sessions or whatnot 🙂</p>\n<p>Hope this helps! 🙌</p>",
      "rawMarkdown": "The reason for the shorter sessions is explained [here](https://www.kaggle.com/competitions/otto-recommender-system/discussion/363554#2015486) by the organizer.\n\nI don't think you can do too much about inferring things about the user as there are no user ids! But maybe you can come up with some smart clustering approach to the sessions or whatnot 🙂\n\nHope this helps! 🙌",
      "votes": null
    },
    {
      "id": "2016486",
      "postDate": "11/04/2022 04:14:08",
      "content": "<p>Thanks! That's very clear now!</p>",
      "rawMarkdown": "Thanks! That's very clear now!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2016236,
      "author_name": "radek1",
      "author_url": "",
      "post_date": "11/03/2022 21:11:27",
      "content": "<p>The reason for the shorter sessions is explained <a href=\"https://www.kaggle.com/competitions/otto-recommender-system/discussion/363554#2015486\" target=\"_blank\">here</a> by the organizer.</p>\n<p>I don't think you can do too much about inferring things about the user as there are no user ids! But maybe you can come up with some smart clustering approach to the sessions or whatnot 🙂</p>\n<p>Hope this helps! 🙌</p>",
      "votes": null,
      "replies": [
        {
          "id": 2016486,
          "author_name": "samsonfha",
          "author_url": "",
          "post_date": "11/04/2022 04:14:08",
          "content": "<p>Thanks! That's very clear now!</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2016034": "I find that the amount of data(actions) in each test session is relatively less(with a large proportion containing only 1 or 2 actions) compared to that in a training session, so it's difficult to identify the feature of the user itself. Does it mean we can hardly generate a label for a certain type of user,  but should focus more on generating labels for the goods/aids?",
    "2016236": "The reason for the shorter sessions is explained [here](https://www.kaggle.com/competitions/otto-recommender-system/discussion/363554#2015486) by the organizer.\n\nI don't think you can do too much about inferring things about the user as there are no user ids! But maybe you can come up with some smart clustering approach to the sessions or whatnot 🙂\n\nHope this helps! 🙌",
    "2016486": "Thanks! That's very clear now!"
  },
  "source": "meta"
}