{
  "id": 1852,
  "title": "Can we use data in test.txt when building models ?",
  "url": "/competitions/kddcup2012-track2/discussion/1852",
  "author_name": "B Yang",
  "post_date": "2012-05-03T22:15:49.050000",
  "votes": 0,
  "comment_count": 0,
  "views": 1536,
  "content": "<p>Can we use the data in test.txt when building models ? For example, let's say I want to rank the ads by the&nbsp;total number of users they were displayed to, can I include the users in test.txt ?</p>\r\n<p>Similarly, when making predictions can we use data in test.txt other than the row we're predicting for ? For example when predicting one instance for user 490234, can I make use of the fact that this user appeared&nbsp;4&nbsp;other&nbsp;times in test.txt, and the settings\r\n associated with those 4&nbsp;other instances&nbsp;?</p>\r\n<p>Presumably in the real production system, test.txt represents future data which you don't have, and what'll be in test.txt depends on your model predictions.</p>",
  "messages": [
    {
      "id": 10783,
      "postDate": "2012-05-03T22:15:49.050Z",
      "content": "<p>Can we use the data in test.txt when building models ? For example, let's say I want to rank the ads by the&nbsp;total number of users they were displayed to, can I include the users in test.txt ?</p>\r\n<p>Similarly, when making predictions can we use data in test.txt other than the row we're predicting for ? For example when predicting one instance for user 490234, can I make use of the fact that this user appeared&nbsp;4&nbsp;other&nbsp;times in test.txt, and the settings\r\n associated with those 4&nbsp;other instances&nbsp;?</p>\r\n<p>Presumably in the real production system, test.txt represents future data which you don't have, and what'll be in test.txt depends on your model predictions.</p>"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "10783": ""
  }
}