{
  "id": 52675,
  "title": "Thoughts about the encoded IPs",
  "url": "/competitions/talkingdata-adtracking-fraud-detection/discussion/52675",
  "author_name": "abandon_learning",
  "post_date": "2018-03-22T00:15:33.295000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>IP address contains signals like Geo information, which is super helpful to understand the users  behaviors of different regions. </p>\n\n<p>I saw two posts:\n<a href=\"https://www.kaggle.com/c/talkingdata-adtracking-fraud-detection/discussion/52374\">https://www.kaggle.com/c/talkingdata-adtracking-fraud-detection/discussion/52374</a>\n<a href=\"https://www.kaggle.com/yuliagm/be-careful-about-ips-as-a-signal\">https://www.kaggle.com/yuliagm/be-careful-about-ips-as-a-signal</a></p>\n\n<p>As Yulia's points out:\n\"Overall the number of IPs (test OR train): 333168\nNumber of IPs that are in both (test AND train): 38164\nNumber of IPs that are in Train and NOT in Test: 239232\nNumber of IPs that are in Test and NOT in Train: 55772\"</p>\n\n<p>Can we confirm that whether IPs are hashed in a logical and consistent way? Otherwise, I believe IPs exists in testing data but not in training will lose a lot of predicting power. If there is any similar post or question like this, please let me know and I will delete this one.</p>",
  "messages": [
    {
      "id": 300819,
      "postDate": "2018-03-22T00:15:33.297Z",
      "content": "<p>IP address contains signals like Geo information, which is super helpful to understand the users  behaviors of different regions. </p>\n\n<p>I saw two posts:\n<a href=\"https://www.kaggle.com/c/talkingdata-adtracking-fraud-detection/discussion/52374\">https://www.kaggle.com/c/talkingdata-adtracking-fraud-detection/discussion/52374</a>\n<a href=\"https://www.kaggle.com/yuliagm/be-careful-about-ips-as-a-signal\">https://www.kaggle.com/yuliagm/be-careful-about-ips-as-a-signal</a></p>\n\n<p>As Yulia's points out:\n\"Overall the number of IPs (test OR train): 333168\nNumber of IPs that are in both (test AND train): 38164\nNumber of IPs that are in Train and NOT in Test: 239232\nNumber of IPs that are in Test and NOT in Train: 55772\"</p>\n\n<p>Can we confirm that whether IPs are hashed in a logical and consistent way? Otherwise, I believe IPs exists in testing data but not in training will lose a lot of predicting power. If there is any similar post or question like this, please let me know and I will delete this one.</p>",
      "rawMarkdown": "IP address contains signals like Geo information, which is super helpful to understand the users  behaviors of different regions. \n\nI saw two posts:\nhttps://www.kaggle.com/c/talkingdata-adtracking-fraud-detection/discussion/52374\nhttps://www.kaggle.com/yuliagm/be-careful-about-ips-as-a-signal\n\nAs Yulia's points out:\n\"Overall the number of IPs (test OR train): 333168\nNumber of IPs that are in both (test AND train): 38164\nNumber of IPs that are in Train and NOT in Test: 239232\nNumber of IPs that are in Test and NOT in Train: 55772\"\n\nCan we confirm that whether IPs are hashed in a logical and consistent way? Otherwise, I believe IPs exists in testing data but not in training will lose a lot of predicting power. If there is any similar post or question like this, please let me know and I will delete this one."
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "300819": "IP address contains signals like Geo information, which is super helpful to understand the users  behaviors of different regions. \n\nI saw two posts:\nhttps://www.kaggle.com/c/talkingdata-adtracking-fraud-detection/discussion/52374\nhttps://www.kaggle.com/yuliagm/be-careful-about-ips-as-a-signal\n\nAs Yulia's points out:\n\"Overall the number of IPs (test OR train): 333168\nNumber of IPs that are in both (test AND train): 38164\nNumber of IPs that are in Train and NOT in Test: 239232\nNumber of IPs that are in Test and NOT in Train: 55772\"\n\nCan we confirm that whether IPs are hashed in a logical and consistent way? Otherwise, I believe IPs exists in testing data but not in training will lose a lot of predicting power. If there is any similar post or question like this, please let me know and I will delete this one."
  }
}