{
  "id": 176402,
  "title": "Why test data does not have all the columns of train data?",
  "url": "/competitions/jigsaw-multilingual-toxic-comment-classification/discussion/176402",
  "author_name": "",
  "post_date": "2020-08-21T15:36:51.316057Z",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Train data have so many columns but test data doesn't why?</p>",
  "messages": [
    {
      "id": "980439",
      "postDate": "08/21/2020 15:36:51",
      "content": "<p>Train data have so many columns but test data doesn't why?</p>",
      "rawMarkdown": "Train data have so many columns but test data doesn't why?",
      "votes": null
    },
    {
      "id": "980850",
      "postDate": "08/21/2020 23:40:49",
      "content": "<p><a href=\"https://www.kaggle.com/harshvardhananand\" target=\"_blank\">@harshvardhananand</a> it depends on the competition, sometimes you can take those variables from the aux tables that they provide or even extract the parameters from the training datasets if there is some kind of ID that can be used to merge the information.</p>\n<p>Sometime that data will not be available at prediction time, and you need to be able to predict with what you have available. I will recommend studying some notebook of top submission to understand what could be the case.</p>",
      "rawMarkdown": "harshvardhananand it depends on the competition, sometimes you can take those variables from the aux tables that they provide or even extract the parameters from the training datasets if there is some kind of ID that can be used to merge the information.\n\nSometime that data will not be available at prediction time, and you need to be able to predict with what you have available. I will recommend studying some notebook of top submission to understand what could be the case.",
      "votes": null
    },
    {
      "id": "981221",
      "postDate": "08/22/2020 09:38:14",
      "content": "<p>Thank you for explanation <a href=\"https://www.kaggle.com/cv13j0\" target=\"_blank\">@cv13j0</a> </p>",
      "rawMarkdown": "Thank you for explanation @cv13j0",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 980850,
      "author_name": "cv13j0",
      "author_url": "",
      "post_date": "08/21/2020 23:40:49",
      "content": "<p><a href=\"https://www.kaggle.com/harshvardhananand\" target=\"_blank\">@harshvardhananand</a> it depends on the competition, sometimes you can take those variables from the aux tables that they provide or even extract the parameters from the training datasets if there is some kind of ID that can be used to merge the information.</p>\n<p>Sometime that data will not be available at prediction time, and you need to be able to predict with what you have available. I will recommend studying some notebook of top submission to understand what could be the case.</p>",
      "votes": null,
      "replies": [
        {
          "id": 981221,
          "author_name": "harshvardhananand",
          "author_url": "",
          "post_date": "08/22/2020 09:38:14",
          "content": "<p>Thank you for explanation <a href=\"https://www.kaggle.com/cv13j0\" target=\"_blank\">@cv13j0</a> </p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "980439": "Train data have so many columns but test data doesn't why?",
    "980850": "harshvardhananand it depends on the competition, sometimes you can take those variables from the aux tables that they provide or even extract the parameters from the training datasets if there is some kind of ID that can be used to merge the information.\n\nSometime that data will not be available at prediction time, and you need to be able to predict with what you have available. I will recommend studying some notebook of top submission to understand what could be the case.",
    "981221": "Thank you for explanation @cv13j0"
  },
  "source": "meta"
}