{
  "id": 14432,
  "title": "Marked as robot based on just one bid?",
  "url": "/competitions/facebook-recruiting-iv-human-or-bot/discussion/14432",
  "author_name": "",
  "post_date": "2015-05-29T08:53:07.520Z",
  "votes": -1,
  "comment_count": 4,
  "views": 1387,
  "content": "<p>There are users that made only one bid. Some of them are marked as robot in training set.&nbsp;</p>\n<p>Don't we need more than observation to judge the behavior of users?</p>",
  "messages": [
    {
      "id": "80348",
      "postDate": "05/29/2015 08:53:07",
      "content": "<p>There are users that made only one bid. Some of them are marked as robot in training set.&nbsp;</p>\n<p>Don't we need more than observation to judge the behavior of users?</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "80353",
      "postDate": "05/29/2015 10:28:20",
      "content": "<p>It's quite likely that the bids data is only an excerpt of the full available data, and the original bot detection algorithm was run on the full bids data. So these users would have suspicious activities in other time periods which we cannot observe, but have only one or a few bids in the time period which we do have access to.</p>\n<p>Unfortunately, that leaves us with some cases where we don't have much to go on.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "81205",
      "postDate": "06/08/2015 05:28:04",
      "content": "<p>I can't aggree more with cfh's opnion.So my trick is to ignore the items in test.csv which made less than 100 bids in bids.csv and the prediction-value of these items are set to 0 as I do believe the robots are just a minority in practice.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "81206",
      "postDate": "06/08/2015 05:32:23",
      "content": "<p>I can't aggree more with cfh's opnion.So my trick is to ignore the items in test.csv which made less than 100 bids in bids.csv and the prediction-value of these items are set to 0 as I do believe the robots are just a minority in practice.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "81245",
      "postDate": "06/08/2015 17:08:01",
      "content": "<p>I also agree with cfh however that maks me wonder: what would the arbitrary&nbsp;algorithm used for this?</p>\n<p>One of my guesses on &quot;labeled as robot by only on bid&quot; is that might be judged by those attributes on which the bid transactions are sparse, e.g. IP, device, and the ratio of robot bids are higher than the standard ration 0.05. I actually engineered a few features&nbsp;based on that hypothesis which turned out work pretty fine.</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 80353,
      "author_name": "cfhlearns",
      "author_url": "",
      "post_date": "05/29/2015 10:28:20",
      "content": "<p>It's quite likely that the bids data is only an excerpt of the full available data, and the original bot detection algorithm was run on the full bids data. So these users would have suspicious activities in other time periods which we cannot observe, but have only one or a few bids in the time period which we do have access to.</p>\n<p>Unfortunately, that leaves us with some cases where we don't have much to go on.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 81205,
      "author_name": "ab19026",
      "author_url": "",
      "post_date": "06/08/2015 05:28:04",
      "content": "<p>I can't aggree more with cfh's opnion.So my trick is to ignore the items in test.csv which made less than 100 bids in bids.csv and the prediction-value of these items are set to 0 as I do believe the robots are just a minority in practice.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 81206,
      "author_name": "ab19026",
      "author_url": "",
      "post_date": "06/08/2015 05:32:23",
      "content": "<p>I can't aggree more with cfh's opnion.So my trick is to ignore the items in test.csv which made less than 100 bids in bids.csv and the prediction-value of these items are set to 0 as I do believe the robots are just a minority in practice.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 81245,
      "author_name": "chriscc",
      "author_url": "",
      "post_date": "06/08/2015 17:08:01",
      "content": "<p>I also agree with cfh however that maks me wonder: what would the arbitrary&nbsp;algorithm used for this?</p>\n<p>One of my guesses on &quot;labeled as robot by only on bid&quot; is that might be judged by those attributes on which the bid transactions are sparse, e.g. IP, device, and the ratio of robot bids are higher than the standard ration 0.05. I actually engineered a few features&nbsp;based on that hypothesis which turned out work pretty fine.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "80348": "",
    "80353": "",
    "81205": "",
    "81206": "",
    "81245": ""
  },
  "source": "meta"
}