{
  "id": 209604,
  "title": "Single lgb, 0.794, 23 features",
  "url": "/competitions/riiid-test-answer-prediction/discussion/209604",
  "author_name": "",
  "post_date": "2021-01-08T01:34:57.275849400Z",
  "votes": 28,
  "comment_count": 8,
  "views": 0,
  "content": "<p>Hi!<br>\nThis is my first Kaggle competition.<br>\nMaybe, I don't have high score, but I want to share my solution and modified for-loop. I update features after task_container_id changes and it gives me slight improvement of score.</p>\n<p>Some magic features:</p>\n<ol>\n<li>Order of task containers + 0.001</li>\n<li>Average difficulty of user questions(content correctness for user) + 0.002</li>\n<li>Average difficulty of user questions per category(content correctness for user per category) + 0.002</li>\n<li>Last 20 answers user accuracy, last 20 answers mean difficulty + 0.001 for each of features<br>\nAlso I tried min/max accuracy per period, but did not have time to add them to the model</li>\n</ol>\n<p>Notebook for training <a href=\"https://www.kaggle.com/fredegrec/training-single-lgb-0-794\" target=\"_blank\">https://www.kaggle.com/fredegrec/training-single-lgb-0-794</a></p>",
  "messages": [
    {
      "id": "1143645",
      "postDate": "01/08/2021 01:34:57",
      "content": "<p>Hi!<br>\nThis is my first Kaggle competition.<br>\nMaybe, I don't have high score, but I want to share my solution and modified for-loop. I update features after task_container_id changes and it gives me slight improvement of score.</p>\n<p>Some magic features:</p>\n<ol>\n<li>Order of task containers + 0.001</li>\n<li>Average difficulty of user questions(content correctness for user) + 0.002</li>\n<li>Average difficulty of user questions per category(content correctness for user per category) + 0.002</li>\n<li>Last 20 answers user accuracy, last 20 answers mean difficulty + 0.001 for each of features<br>\nAlso I tried min/max accuracy per period, but did not have time to add them to the model</li>\n</ol>\n<p>Notebook for training <a href=\"https://www.kaggle.com/fredegrec/training-single-lgb-0-794\" target=\"_blank\">https://www.kaggle.com/fredegrec/training-single-lgb-0-794</a></p>",
      "rawMarkdown": "Hi!\nThis is my first Kaggle competition.\nMaybe, I don't have high score, but I want to share my solution and modified for-loop. I update features after task_container_id changes and it gives me slight improvement of score.\n\nSome magic features:\n1. Order of task containers + 0.001\n2. Average difficulty of user questions(content correctness for user) + 0.002\n3. Average difficulty of user questions per category(content correctness for user per category) + 0.002\n4. Last 20 answers user accuracy, last 20 answers mean difficulty + 0.001 for each of features\nAlso I tried min/max accuracy per period, but did not have time to add them to the model\n\nNotebook for training https://www.kaggle.com/fredegrec/training-single-lgb-0-794",
      "votes": null
    },
    {
      "id": "1143649",
      "postDate": "01/08/2021 01:40:33",
      "content": "<p>magic!! just 23 features </p>",
      "rawMarkdown": "magic!! just 23 features",
      "votes": null
    },
    {
      "id": "1143658",
      "postDate": "01/08/2021 01:48:44",
      "content": "<p>Great job, I<code>m wondering what does \"Order of task containers\" mean, what is this feature</code>s name in your notebook?</p>",
      "rawMarkdown": "Great job, I`m wondering what does \"Order of task containers\" mean, what is this feature`s name in your notebook?",
      "votes": null
    },
    {
      "id": "1143663",
      "postDate": "01/08/2021 01:51:04",
      "content": "<p>It is 'wrong_container_u'<br>\nThanks!</p>",
      "rawMarkdown": "It is 'wrong_container_u'\nThanks!",
      "votes": null
    },
    {
      "id": "1143666",
      "postDate": "01/08/2021 01:54:39",
      "content": "<p>Compare for each user the number of the previous container and the current one, if it is greater than current. Update feature on new container</p>",
      "rawMarkdown": "Compare for each user the number of the previous container and the current one, if it is greater than current. Update feature on new container",
      "votes": null
    },
    {
      "id": "1143678",
      "postDate": "01/08/2021 02:04:03",
      "content": "<p>what  a excellent job, my team's lgb got about 784 on the lb.</p>",
      "rawMarkdown": "what  a excellent job, my team's lgb got about 784 on the lb.",
      "votes": null
    },
    {
      "id": "1143703",
      "postDate": "01/08/2021 02:33:17",
      "content": "<p>Great work! Thank you for sharing your notebook.</p>",
      "rawMarkdown": "Great work! Thank you for sharing your notebook.",
      "votes": null
    },
    {
      "id": "1143758",
      "postDate": "01/08/2021 03:40:47",
      "content": "<p>0.794 with 23 features is very impressive <a href=\"https://www.kaggle.com/fredegrec\" target=\"_blank\">@fredegrec</a>! Awesome work &amp; congratulations!</p>",
      "rawMarkdown": "0.794 with 23 features is very impressive @fredegrec! Awesome work & congratulations!",
      "votes": null
    },
    {
      "id": "1144752",
      "postDate": "01/08/2021 16:42:01",
      "content": "<p>thanks for sharing! unbelievable!</p>",
      "rawMarkdown": "thanks for sharing! unbelievable!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1143649,
      "author_name": "yangxiaoshuai",
      "author_url": "",
      "post_date": "01/08/2021 01:40:33",
      "content": "<p>magic!! just 23 features </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1143658,
      "author_name": "superchenhao",
      "author_url": "",
      "post_date": "01/08/2021 01:48:44",
      "content": "<p>Great job, I<code>m wondering what does \"Order of task containers\" mean, what is this feature</code>s name in your notebook?</p>",
      "votes": null,
      "replies": [
        {
          "id": 1143663,
          "author_name": "fredegrec",
          "author_url": "",
          "post_date": "01/08/2021 01:51:04",
          "content": "<p>It is 'wrong_container_u'<br>\nThanks!</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1143666,
          "author_name": "fredegrec",
          "author_url": "",
          "post_date": "01/08/2021 01:54:39",
          "content": "<p>Compare for each user the number of the previous container and the current one, if it is greater than current. Update feature on new container</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1143678,
      "author_name": "luffy521",
      "author_url": "",
      "post_date": "01/08/2021 02:04:03",
      "content": "<p>what  a excellent job, my team's lgb got about 784 on the lb.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1143703,
      "author_name": "sishihara",
      "author_url": "",
      "post_date": "01/08/2021 02:33:17",
      "content": "<p>Great work! Thank you for sharing your notebook.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1143758,
      "author_name": "doctorkael",
      "author_url": "",
      "post_date": "01/08/2021 03:40:47",
      "content": "<p>0.794 with 23 features is very impressive <a href=\"https://www.kaggle.com/fredegrec\" target=\"_blank\">@fredegrec</a>! Awesome work &amp; congratulations!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1144752,
      "author_name": "zjjszj2",
      "author_url": "",
      "post_date": "01/08/2021 16:42:01",
      "content": "<p>thanks for sharing! unbelievable!</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1143645": "Hi!\nThis is my first Kaggle competition.\nMaybe, I don't have high score, but I want to share my solution and modified for-loop. I update features after task_container_id changes and it gives me slight improvement of score.\n\nSome magic features:\n1. Order of task containers + 0.001\n2. Average difficulty of user questions(content correctness for user) + 0.002\n3. Average difficulty of user questions per category(content correctness for user per category) + 0.002\n4. Last 20 answers user accuracy, last 20 answers mean difficulty + 0.001 for each of features\nAlso I tried min/max accuracy per period, but did not have time to add them to the model\n\nNotebook for training https://www.kaggle.com/fredegrec/training-single-lgb-0-794",
    "1143649": "magic!! just 23 features",
    "1143658": "Great job, I`m wondering what does \"Order of task containers\" mean, what is this feature`s name in your notebook?",
    "1143663": "It is 'wrong_container_u'\nThanks!",
    "1143666": "Compare for each user the number of the previous container and the current one, if it is greater than current. Update feature on new container",
    "1143678": "what  a excellent job, my team's lgb got about 784 on the lb.",
    "1143703": "Great work! Thank you for sharing your notebook.",
    "1143758": "0.794 with 23 features is very impressive @fredegrec! Awesome work & congratulations!",
    "1144752": "thanks for sharing! unbelievable!"
  },
  "source": "meta"
}