{
  "id": 79591,
  "title": "Month Lag and Reference Date",
  "url": "/competitions/elo-merchant-category-recommendation/discussion/79591",
  "author_name": "",
  "post_date": "2019-02-05T20:17:50.109933Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Digging a bit deep with just bare raw data it seems the reference date for the historical transactions and new transactions data is different.</p>\n\n<p><strong>For Historical Transactions:</strong>\nFeb 2018 seems to be the reference date and the month lag then should refers to how stale the data is for that particular card_Id.</p>\n\n<p><strong>For New Merchant Transactions:</strong>\nThe reference date which is actually the month of last purchase date by the card_id. \nThe problem statement mentions the time period of 2 months, that would mean only the outlier transactions that have been done 2 months beyond the last transaction has been taken in account. This makes sense when we find only values of 1 and 2 in the month lag. So if anyone picks up the last transaction date from Historical transactions and to the month lag that gives the purchase date(precise to the month) for that entry. (this also makes sense for the negative sign on month lag in new transactions.)</p>\n\n<p>Tried my theory on a few cards and it made sense.\nAlthough this knowledge didn't help my model at all, hope someone is able to make better use of it.</p>",
  "messages": [
    {
      "id": "466688",
      "postDate": "02/05/2019 20:17:50",
      "content": "<p>Digging a bit deep with just bare raw data it seems the reference date for the historical transactions and new transactions data is different.</p>\n\n<p><strong>For Historical Transactions:</strong>\nFeb 2018 seems to be the reference date and the month lag then should refers to how stale the data is for that particular card_Id.</p>\n\n<p><strong>For New Merchant Transactions:</strong>\nThe reference date which is actually the month of last purchase date by the card_id. \nThe problem statement mentions the time period of 2 months, that would mean only the outlier transactions that have been done 2 months beyond the last transaction has been taken in account. This makes sense when we find only values of 1 and 2 in the month lag. So if anyone picks up the last transaction date from Historical transactions and to the month lag that gives the purchase date(precise to the month) for that entry. (this also makes sense for the negative sign on month lag in new transactions.)</p>\n\n<p>Tried my theory on a few cards and it made sense.\nAlthough this knowledge didn't help my model at all, hope someone is able to make better use of it.</p>",
      "rawMarkdown": "Digging a bit deep with just bare raw data it seems the reference date for the historical transactions and new transactions data is different.\n\n**For Historical Transactions:**\nFeb 2018 seems to be the reference date and the month lag then should refers to how stale the data is for that particular card_Id.\n\n**For New Merchant Transactions:**\nThe reference date which is actually the month of last purchase date by the card_id. \nThe problem statement mentions the time period of 2 months, that would mean only the outlier transactions that have been done 2 months beyond the last transaction has been taken in account. This makes sense when we find only values of 1 and 2 in the month lag. So if anyone picks up the last transaction date from Historical transactions and to the month lag that gives the purchase date(precise to the month) for that entry. (this also makes sense for the negative sign on month lag in new transactions.)\n\nTried my theory on a few cards and it made sense.\nAlthough this knowledge didn't help my model at all, hope someone is able to make better use of it.",
      "votes": null
    },
    {
      "id": "466748",
      "postDate": "02/05/2019 22:11:59",
      "content": "<p>Good point. </p>",
      "rawMarkdown": "Good point.",
      "votes": null
    },
    {
      "id": "466874",
      "postDate": "02/06/2019 05:56:26",
      "content": "<p>I think the reference date is actually the first purchase by the card_id for that merchant_id.  </p>",
      "rawMarkdown": "I think the reference date is actually the first purchase by the card_id for that merchant_id.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 466748,
      "author_name": "anystone",
      "author_url": "",
      "post_date": "02/05/2019 22:11:59",
      "content": "<p>Good point. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 466874,
      "author_name": "plhirulkar",
      "author_url": "",
      "post_date": "02/06/2019 05:56:26",
      "content": "<p>I think the reference date is actually the first purchase by the card_id for that merchant_id.  </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "466688": "Digging a bit deep with just bare raw data it seems the reference date for the historical transactions and new transactions data is different.\n\n**For Historical Transactions:**\nFeb 2018 seems to be the reference date and the month lag then should refers to how stale the data is for that particular card_Id.\n\n**For New Merchant Transactions:**\nThe reference date which is actually the month of last purchase date by the card_id. \nThe problem statement mentions the time period of 2 months, that would mean only the outlier transactions that have been done 2 months beyond the last transaction has been taken in account. This makes sense when we find only values of 1 and 2 in the month lag. So if anyone picks up the last transaction date from Historical transactions and to the month lag that gives the purchase date(precise to the month) for that entry. (this also makes sense for the negative sign on month lag in new transactions.)\n\nTried my theory on a few cards and it made sense.\nAlthough this knowledge didn't help my model at all, hope someone is able to make better use of it.",
    "466748": "Good point.",
    "466874": "I think the reference date is actually the first purchase by the card_id for that merchant_id."
  },
  "source": "meta"
}