{
  "id": 483011,
  "title": "If WEEK_NUM will be deleted...",
  "url": "/competitions/home-credit-credit-risk-model-stability/discussion/483011",
  "author_name": "",
  "post_date": "2024-03-10T14:38:27.923773Z",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I am the freshman of credit and kaggle, I'm  considering using “WEEK_NUM\" to build features. I want to know if \"WEEK_NUM\" will be deleted in \"test data\", does this mean that I can't build new feature using \"WEEK_NUM\" in the \"train data\"? </p>",
  "messages": [
    {
      "id": "2690463",
      "postDate": "03/10/2024 14:38:27",
      "content": "<p>I am the freshman of credit and kaggle, I'm  considering using “WEEK_NUM\" to build features. I want to know if \"WEEK_NUM\" will be deleted in \"test data\", does this mean that I can't build new feature using \"WEEK_NUM\" in the \"train data\"? </p>",
      "rawMarkdown": "I am the freshman of credit and kaggle, I'm  considering using “WEEK_NUM\" to build features. I want to know if \"WEEK_NUM\" will be deleted in \"test data\", does this mean that I can't build new feature using \"WEEK_NUM\" in the \"train data\"?",
      "votes": null
    },
    {
      "id": "2690544",
      "postDate": "03/10/2024 16:30:52",
      "content": "<p>Check out <a href=\"https://www.kaggle.com/competitions/home-credit-credit-risk-model-stability/discussion/482474\" target=\"_blank\">https://www.kaggle.com/competitions/home-credit-credit-risk-model-stability/discussion/482474</a>, more specifically:</p>\n<blockquote>\n  <p>As we indicated previously, we will be changing date columns in test_base table, namely date_decision, MONTH and WEEK_NUM. The MONTH and WEEK_NUM columns will now only have one constant value.</p>\n</blockquote>\n<p>The answer to your question is that using <code>WEEK_NUM</code> as a feature won't generalize to the test set because there this column will have a unique placeholder value with no (valuable) predictive information.</p>",
      "rawMarkdown": "Check out https://www.kaggle.com/competitions/home-credit-credit-risk-model-stability/discussion/482474, more specifically:\n>As we indicated previously, we will be changing date columns in test_base table, namely date_decision, MONTH and WEEK_NUM. The MONTH and WEEK_NUM columns will now only have one constant value.\n\nThe answer to your question is that using `WEEK_NUM` as a feature won't generalize to the test set because there this column will have a unique placeholder value with no (valuable) predictive information.",
      "votes": null
    },
    {
      "id": "2691063",
      "postDate": "03/11/2024 02:44:17",
      "content": "<p>Thank you for your answer</p>",
      "rawMarkdown": "Thank you for your answer",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2690544,
      "author_name": "asajatovic",
      "author_url": "",
      "post_date": "03/10/2024 16:30:52",
      "content": "<p>Check out <a href=\"https://www.kaggle.com/competitions/home-credit-credit-risk-model-stability/discussion/482474\" target=\"_blank\">https://www.kaggle.com/competitions/home-credit-credit-risk-model-stability/discussion/482474</a>, more specifically:</p>\n<blockquote>\n  <p>As we indicated previously, we will be changing date columns in test_base table, namely date_decision, MONTH and WEEK_NUM. The MONTH and WEEK_NUM columns will now only have one constant value.</p>\n</blockquote>\n<p>The answer to your question is that using <code>WEEK_NUM</code> as a feature won't generalize to the test set because there this column will have a unique placeholder value with no (valuable) predictive information.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2691063,
          "author_name": "shijilong",
          "author_url": "",
          "post_date": "03/11/2024 02:44:17",
          "content": "<p>Thank you for your answer</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2690463": "I am the freshman of credit and kaggle, I'm  considering using “WEEK_NUM\" to build features. I want to know if \"WEEK_NUM\" will be deleted in \"test data\", does this mean that I can't build new feature using \"WEEK_NUM\" in the \"train data\"?",
    "2690544": "Check out https://www.kaggle.com/competitions/home-credit-credit-risk-model-stability/discussion/482474, more specifically:\n>As we indicated previously, we will be changing date columns in test_base table, namely date_decision, MONTH and WEEK_NUM. The MONTH and WEEK_NUM columns will now only have one constant value.\n\nThe answer to your question is that using `WEEK_NUM` as a feature won't generalize to the test set because there this column will have a unique placeholder value with no (valuable) predictive information.",
    "2691063": "Thank you for your answer"
  },
  "source": "meta"
}