{
  "id": 320719,
  "title": "Are you using postal code?",
  "url": "/competitions/h-and-m-personalized-fashion-recommendations/discussion/320719",
  "author_name": "",
  "post_date": "2022-04-23T06:12:29.437227700Z",
  "votes": 5,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I think <code>postal code</code> is not useful because it is hashed and customers have so different postal code.</p>\n<p>Are you using <code>postal code</code> in modeling? Can you share how are you using?</p>",
  "messages": [
    {
      "id": "1765024",
      "postDate": "04/23/2022 06:12:29",
      "content": "<p>I think <code>postal code</code> is not useful because it is hashed and customers have so different postal code.</p>\n<p>Are you using <code>postal code</code> in modeling? Can you share how are you using?</p>",
      "rawMarkdown": "I think `postal code` is not useful because it is hashed and customers have so different postal code.\n\nAre you using `postal code` in modeling? Can you share how are you using?",
      "votes": null
    },
    {
      "id": "1765117",
      "postDate": "04/23/2022 07:56:48",
      "content": "<p>I don't use postal code.</p>",
      "rawMarkdown": "I don't use postal code.",
      "votes": null
    },
    {
      "id": "1784942",
      "postDate": "05/11/2022 15:59:09",
      "content": "<p>Yes. I tried to use it to predict future purchases of cold-start users; i.e., customers with no transaction data.<br>\nI found that <code>postal code</code> and <code>sales channel id</code> are correlated in the sense that customers having a particular <code>postal code</code> are more likely to buy products in stores / via website; e.g. around 120,000 customers in the largest postal code group (probably those who didn't register their postal code) tend to use stores. So I used the average <code>sales channel id</code> of each postal code to guess which sales channel each cold-start customer would use.</p>",
      "rawMarkdown": "Yes. I tried to use it to predict future purchases of cold-start users; i.e., customers with no transaction data.\nI found that `postal code` and `sales channel id` are correlated in the sense that customers having a particular `postal code` are more likely to buy products in stores / via website; e.g. around 120,000 customers in the largest postal code group (probably those who didn't register their postal code) tend to use stores. So I used the average `sales channel id` of each postal code to guess which sales channel each cold-start customer would use.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1765117,
      "author_name": "adldotori",
      "author_url": "",
      "post_date": "04/23/2022 07:56:48",
      "content": "<p>I don't use postal code.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1784942,
      "author_name": "negoto",
      "author_url": "",
      "post_date": "05/11/2022 15:59:09",
      "content": "<p>Yes. I tried to use it to predict future purchases of cold-start users; i.e., customers with no transaction data.<br>\nI found that <code>postal code</code> and <code>sales channel id</code> are correlated in the sense that customers having a particular <code>postal code</code> are more likely to buy products in stores / via website; e.g. around 120,000 customers in the largest postal code group (probably those who didn't register their postal code) tend to use stores. So I used the average <code>sales channel id</code> of each postal code to guess which sales channel each cold-start customer would use.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1765024": "I think `postal code` is not useful because it is hashed and customers have so different postal code.\n\nAre you using `postal code` in modeling? Can you share how are you using?",
    "1765117": "I don't use postal code.",
    "1784942": "Yes. I tried to use it to predict future purchases of cold-start users; i.e., customers with no transaction data.\nI found that `postal code` and `sales channel id` are correlated in the sense that customers having a particular `postal code` are more likely to buy products in stores / via website; e.g. around 120,000 customers in the largest postal code group (probably those who didn't register their postal code) tend to use stores. So I used the average `sales channel id` of each postal code to guess which sales channel each cold-start customer would use."
  },
  "source": "meta"
}