{
  "id": 587726,
  "title": "How to solve Recommendation System problems ?",
  "url": "/competitions/aeroclub-recsys-2025/discussion/587726",
  "author_name": "",
  "post_date": "2025-07-02T12:00:21.341620800Z",
  "votes": 2,
  "comment_count": 7,
  "views": 0,
  "content": "<p>I published a discussion few weeks ago regarding solving a problem using recommendation system it can be useful for this competition as it is recommendation problem , the said article can be found <a href=\"https://www.kaggle.com/competitions/playground-series-s5e6/discussion/586376#3233996\" target=\"_blank\">here</a></p>",
  "messages": [
    {
      "id": "3239025",
      "postDate": "07/02/2025 12:00:21",
      "content": "<p>I published a discussion few weeks ago regarding solving a problem using recommendation system it can be useful for this competition as it is recommendation problem , the said article can be found <a href=\"https://www.kaggle.com/competitions/playground-series-s5e6/discussion/586376#3233996\" target=\"_blank\">here</a></p>",
      "rawMarkdown": "I published a discussion few weeks ago regarding solving a problem using recommendation system it can be useful for this competition as it is recommendation problem , the said article can be found [here](https://www.kaggle.com/competitions/playground-series-s5e6/discussion/586376#3233996)",
      "votes": null
    },
    {
      "id": "3239048",
      "postDate": "07/02/2025 12:41:13",
      "content": "<p>Yes, but.. :-).  30% of users without history, 30% of train have just one positive, and more than 50% less that 3 positive.  </p>",
      "rawMarkdown": "Yes, but.. :-).  30% of users without history, 30% of train have just one positive, and more than 50% less that 3 positive.",
      "votes": null
    },
    {
      "id": "3239055",
      "postDate": "07/02/2025 12:51:01",
      "content": "<p>Still if someone wants to give it a shot 😀</p>",
      "rawMarkdown": "Still if someone wants to give it a shot 😀",
      "votes": null
    },
    {
      "id": "3251666",
      "postDate": "07/21/2025 11:08:14",
      "content": "<p>I tried adding some user-related feature engineering, but the effect was very poor</p>",
      "rawMarkdown": "I tried adding some user-related feature engineering, but the effect was very poor",
      "votes": null
    },
    {
      "id": "3251673",
      "postDate": "07/21/2025 11:31:06",
      "content": "<p>However, some recommendation-related methods perform very well on the training set, but poorly on the test set. I think this is a direction for breakthrough.</p>",
      "rawMarkdown": "However, some recommendation-related methods perform very well on the training set, but poorly on the test set. I think this is a direction for breakthrough.",
      "votes": null
    },
    {
      "id": "3251746",
      "postDate": "07/21/2025 13:44:07",
      "content": "<p>yeah, there's just not enough user behavior history…</p>",
      "rawMarkdown": "yeah, there's just not enough user behavior history...",
      "votes": null
    },
    {
      "id": "3251830",
      "postDate": "07/21/2025 16:28:03",
      "content": "<p>yes, agree</p>",
      "rawMarkdown": "yes, agree",
      "votes": null
    },
    {
      "id": "3253330",
      "postDate": "07/24/2025 13:16:23",
      "content": "<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F22034645%2F3a55ed346490daca6da251fd072d316a%2Fkgl.JPG?generation=1753362973799596&amp;alt=media\" alt=\"\">          Summertime, summertime sadness</p>",
      "rawMarkdown": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F22034645%2F3a55ed346490daca6da251fd072d316a%2Fkgl.JPG?generation=1753362973799596&alt=media)          Summertime, summertime sadness",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3239048,
      "author_name": "sergeyqt2024",
      "author_url": "",
      "post_date": "07/02/2025 12:41:13",
      "content": "<p>Yes, but.. :-).  30% of users without history, 30% of train have just one positive, and more than 50% less that 3 positive.  </p>",
      "votes": null,
      "replies": [
        {
          "id": 3239055,
          "author_name": "sayedathar11",
          "author_url": "",
          "post_date": "07/02/2025 12:51:01",
          "content": "<p>Still if someone wants to give it a shot 😀</p>",
          "votes": null,
          "replies": [
            {
              "id": 3251666,
              "author_name": "mengshangyu11",
              "author_url": "",
              "post_date": "07/21/2025 11:08:14",
              "content": "<p>I tried adding some user-related feature engineering, but the effect was very poor</p>",
              "votes": null,
              "replies": [
                {
                  "id": 3251746,
                  "author_name": "mango789",
                  "author_url": "",
                  "post_date": "07/21/2025 13:44:07",
                  "content": "<p>yeah, there's just not enough user behavior history…</p>",
                  "votes": null,
                  "replies": []
                }
              ]
            },
            {
              "id": 3251673,
              "author_name": "mengshangyu11",
              "author_url": "",
              "post_date": "07/21/2025 11:31:06",
              "content": "<p>However, some recommendation-related methods perform very well on the training set, but poorly on the test set. I think this is a direction for breakthrough.</p>",
              "votes": null,
              "replies": [
                {
                  "id": 3251830,
                  "author_name": "sergeyqt2024",
                  "author_url": "",
                  "post_date": "07/21/2025 16:28:03",
                  "content": "<p>yes, agree</p>",
                  "votes": null,
                  "replies": []
                },
                {
                  "id": 3253330,
                  "author_name": "sergeyqt2024",
                  "author_url": "",
                  "post_date": "07/24/2025 13:16:23",
                  "content": "<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F22034645%2F3a55ed346490daca6da251fd072d316a%2Fkgl.JPG?generation=1753362973799596&amp;alt=media\" alt=\"\">          Summertime, summertime sadness</p>",
                  "votes": null,
                  "replies": []
                }
              ]
            }
          ]
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3239025": "I published a discussion few weeks ago regarding solving a problem using recommendation system it can be useful for this competition as it is recommendation problem , the said article can be found [here](https://www.kaggle.com/competitions/playground-series-s5e6/discussion/586376#3233996)",
    "3239048": "Yes, but.. :-).  30% of users without history, 30% of train have just one positive, and more than 50% less that 3 positive.",
    "3239055": "Still if someone wants to give it a shot 😀",
    "3251666": "I tried adding some user-related feature engineering, but the effect was very poor",
    "3251673": "However, some recommendation-related methods perform very well on the training set, but poorly on the test set. I think this is a direction for breakthrough.",
    "3251746": "yeah, there's just not enough user behavior history...",
    "3251830": "yes, agree",
    "3253330": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F22034645%2F3a55ed346490daca6da251fd072d316a%2Fkgl.JPG?generation=1753362973799596&alt=media)          Summertime, summertime sadness"
  },
  "source": "meta"
}