{
  "id": 57792,
  "title": "What is your strategy to win?",
  "url": "/competitions/avito-demand-prediction/discussion/57792",
  "author_name": "Nomoreaday",
  "post_date": "2018-05-29T07:04:51.697000",
  "votes": 0,
  "comment_count": 11,
  "views": 0,
  "content": "<p>Please give me some advice.</p>",
  "messages": [
    {
      "id": 335182,
      "postDate": "2018-05-29T10:50:43.283Z",
      "content": "<p><img src=\"https://qph.ec.quoracdn.net/main-qimg-5459cc0de288053771a5d7f66115c8a7.webp\" alt=\"exodia\"></p>",
      "rawMarkdown": "![exodia][1]\n\n\n  [1]: https://qph.ec.quoracdn.net/main-qimg-5459cc0de288053771a5d7f66115c8a7.webp",
      "votes": 11
    },
    {
      "id": 335248,
      "postDate": "2018-05-29T12:42:33.810Z",
      "content": "<p>Going deeper</p>",
      "rawMarkdown": "Going deeper",
      "votes": 3,
      "replies": [
        {
          "id": 336328,
          "postDate": "2018-05-31T12:50:52.347Z",
          "content": "<p>Dieter, please!</p>",
          "rawMarkdown": "Dieter, please!"
        },
        {
          "id": 337279,
          "postDate": "2018-06-02T12:43:51.057Z",
          "content": "<p>more seriously:</p>\n\n<p>building good lgb, xgb and NN models and stacking them</p>",
          "rawMarkdown": "more seriously:\n\nbuilding good lgb, xgb and NN models and stacking them"
        }
      ]
    },
    {
      "id": 335122,
      "postDate": "2018-05-29T07:52:24.957Z",
      "content": "<p>I find it hard to improve lb score, by adding group features. </p>",
      "rawMarkdown": "I find it hard to improve lb score, by adding group features. ",
      "replies": [
        {
          "id": 335184,
          "postDate": "2018-05-29T10:52:04.090Z",
          "content": "<p>Because everyone does that</p>",
          "rawMarkdown": "Because everyone does that",
          "votes": 1
        },
        {
          "id": 335199,
          "postDate": "2018-05-29T11:12:29.167Z",
          "content": "<p>And It seem that I have to do it harder, because everyone does - -</p>",
          "rawMarkdown": "And It seem that I have to do it harder, because everyone does - -"
        },
        {
          "id": 335218,
          "postDate": "2018-05-29T11:50:48.097Z",
          "content": "<p>I tried multiple combination of features and couldn't go past the bottleneck score. Looks like we need to look for something else, may be the magic features. Also need to check on the usage of Image data, active data and periods data.</p>",
          "rawMarkdown": "I tried multiple combination of features and couldn't go past the bottleneck score. Looks like we need to look for something else, may be the magic features. Also need to check on the usage of Image data, active data and periods data."
        },
        {
          "id": 335221,
          "postDate": "2018-05-29T11:52:59.310Z",
          "content": "<p>Using active for data augmentation maybe</p>",
          "rawMarkdown": "Using active for data augmentation maybe",
          "votes": 2
        },
        {
          "id": 340675,
          "postDate": "2018-06-10T01:09:40.577Z",
          "content": "<p>Has anyone explored deeper using <code>param_1:param_3</code>, etc.? Not all parameters are related to one another. For example, some parameters may contain a car brand while others contain a shoe size. I guess built in methods like categoricals help, but keeping them as a numeric wouldn't make sense. Has anyone seen any value from exploring this further?</p>",
          "rawMarkdown": "Has anyone explored deeper using `param_1:param_3`, etc.? Not all parameters are related to one another. For example, some parameters may contain a car brand while others contain a shoe size. I guess built in methods like categoricals help, but keeping them as a numeric wouldn't make sense. Has anyone seen any value from exploring this further?"
        },
        {
          "id": 340735,
          "postDate": "2018-06-10T06:37:10.603Z",
          "rawMarkdown": "",
          "isDeleted": true
        }
      ]
    },
    {
      "id": 335103,
      "postDate": "2018-05-29T07:04:51.697Z",
      "content": "<p>Please give me some advice.</p>",
      "rawMarkdown": "Please give me some advice."
    }
  ],
  "comments": [
    {
      "id": 335182,
      "author_name": "Adil Zouitine",
      "author_url": "",
      "post_date": "2018-05-29T10:50:43.283000",
      "content": "<p><img src=\"https://qph.ec.quoracdn.net/main-qimg-5459cc0de288053771a5d7f66115c8a7.webp\" alt=\"exodia\"></p>",
      "votes": 11,
      "replies": []
    },
    {
      "id": 335248,
      "author_name": "Dieter",
      "author_url": "",
      "post_date": "2018-05-29T12:42:33.810000",
      "content": "<p>Going deeper</p>",
      "votes": 3,
      "replies": [
        {
          "id": 336328,
          "author_name": "Max Halford",
          "author_url": "",
          "post_date": "2018-05-31T12:50:52.347000",
          "content": "<p>Dieter, please!</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 337279,
          "author_name": "Dieter",
          "author_url": "",
          "post_date": "2018-06-02T12:43:51.057000",
          "content": "<p>more seriously:</p>\n\n<p>building good lgb, xgb and NN models and stacking them</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 335122,
      "author_name": "Johnny Liu",
      "author_url": "",
      "post_date": "2018-05-29T07:52:24.957000",
      "content": "<p>I find it hard to improve lb score, by adding group features. </p>",
      "votes": 0,
      "replies": [
        {
          "id": 335184,
          "author_name": "Adil Zouitine",
          "author_url": "",
          "post_date": "2018-05-29T10:52:04.090000",
          "content": "<p>Because everyone does that</p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 335199,
          "author_name": "Johnny Liu",
          "author_url": "",
          "post_date": "2018-05-29T11:12:29.167000",
          "content": "<p>And It seem that I have to do it harder, because everyone does - -</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 335218,
          "author_name": "Samrat Pandiri",
          "author_url": "",
          "post_date": "2018-05-29T11:50:48.097000",
          "content": "<p>I tried multiple combination of features and couldn't go past the bottleneck score. Looks like we need to look for something else, may be the magic features. Also need to check on the usage of Image data, active data and periods data.</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 335221,
          "author_name": "Adil Zouitine",
          "author_url": "",
          "post_date": "2018-05-29T11:52:59.310000",
          "content": "<p>Using active for data augmentation maybe</p>",
          "votes": 2,
          "replies": []
        },
        {
          "id": 340675,
          "author_name": "Thomas Yokota",
          "author_url": "",
          "post_date": "2018-06-10T01:09:40.577000",
          "content": "<p>Has anyone explored deeper using <code>param_1:param_3</code>, etc.? Not all parameters are related to one another. For example, some parameters may contain a car brand while others contain a shoe size. I guess built in methods like categoricals help, but keeping them as a numeric wouldn't make sense. Has anyone seen any value from exploring this further?</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 340735,
          "author_name": "",
          "author_url": "",
          "post_date": "2018-06-10T06:37:10.603000",
          "content": "",
          "votes": 0,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "335182": "![exodia][1]\n\n\n  [1]: https://qph.ec.quoracdn.net/main-qimg-5459cc0de288053771a5d7f66115c8a7.webp",
    "335248": "Going deeper",
    "335122": "I find it hard to improve lb score, by adding group features. ",
    "335103": "Please give me some advice."
  }
}