{
  "id": 73042,
  "title": "New and Historical Transactions unimportant features",
  "url": "/competitions/elo-merchant-category-recommendation/discussion/73042",
  "author_name": "",
  "post_date": "2018-11-29T07:17:46.559178600Z",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>If we are trying to find consumer loyality, do the following features need to be considered?</p>\n\n<ol>\n<li>state_id</li>\n<li>subsector_id </li>\n<li>installments            </li>\n<li>purchase_date           </li>\n<li>subsector_id            </li>\n<li>merchant_category_id    </li>\n</ol>\n\n<p>At least some of them seems to be just noise</p>",
  "messages": [
    {
      "id": "429681",
      "postDate": "11/29/2018 07:17:46",
      "content": "<p>If we are trying to find consumer loyality, do the following features need to be considered?</p>\n\n<ol>\n<li>state_id</li>\n<li>subsector_id </li>\n<li>installments            </li>\n<li>purchase_date           </li>\n<li>subsector_id            </li>\n<li>merchant_category_id    </li>\n</ol>\n\n<p>At least some of them seems to be just noise</p>",
      "rawMarkdown": "If we are trying to find consumer loyality, do the following features need to be considered?\n\n 1. state_id\n 2. subsector_id \n 3. installments            \n 4. purchase_date           \n 5. subsector_id            \n 6. merchant_category_id    \n\nAt least some of them seems to be just noise",
      "votes": null
    },
    {
      "id": "429813",
      "postDate": "11/29/2018 11:34:11",
      "content": "<p>FWIW, I have managed to extract some useful (= non-zero importance in lgb) features out of each of the ones you mention. Mostly I used descriptive statistics, counting unique values etc. Check the top kernel in LB, there are some nice ideas there.</p>",
      "rawMarkdown": "FWIW, I have managed to extract some useful (= non-zero importance in lgb) features out of each of the ones you mention. Mostly I used descriptive statistics, counting unique values etc. Check the top kernel in LB, there are some nice ideas there.",
      "votes": null
    },
    {
      "id": "465726",
      "postDate": "02/03/2019 21:25:28",
      "content": "<p>Purchase date is definitely an important feature. </p>",
      "rawMarkdown": "Purchase date is definitely an important feature.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 429813,
      "author_name": "konradb",
      "author_url": "",
      "post_date": "11/29/2018 11:34:11",
      "content": "<p>FWIW, I have managed to extract some useful (= non-zero importance in lgb) features out of each of the ones you mention. Mostly I used descriptive statistics, counting unique values etc. Check the top kernel in LB, there are some nice ideas there.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 465726,
      "author_name": "pukkinming",
      "author_url": "",
      "post_date": "02/03/2019 21:25:28",
      "content": "<p>Purchase date is definitely an important feature. </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "429681": "If we are trying to find consumer loyality, do the following features need to be considered?\n\n 1. state_id\n 2. subsector_id \n 3. installments            \n 4. purchase_date           \n 5. subsector_id            \n 6. merchant_category_id    \n\nAt least some of them seems to be just noise",
    "429813": "FWIW, I have managed to extract some useful (= non-zero importance in lgb) features out of each of the ones you mention. Mostly I used descriptive statistics, counting unique values etc. Check the top kernel in LB, there are some nice ideas there.",
    "465726": "Purchase date is definitely an important feature."
  },
  "source": "meta"
}