{
  "id": 3548,
  "title": "my approach",
  "url": "/competitions/flight/discussion/3548",
  "author_name": "",
  "post_date": "2013-01-07T12:43:43.770Z",
  "votes": null,
  "comment_count": 1,
  "views": 2198,
  "content": "<pre>I have tried following approaches with similar results:<br><br>For all I perform cross-validation on the training data by applying same cutoff times as for PublicLeaderboard.<br><br>a) Extract most recent arrival time for runway/gate from flighthistoryevents.csv file and use them:<br>  Gives same score as benchmark<br>  <br>b) Try to predict the arrival time for runway/gate using the following as predictors: expected runway/gate arrival times, scheduled air time, type of events, .....<br>  SLightly better than benchmark<br>  <br>c) Forget the mins since midnight and instead predict flight times<br>  Same as b)<br>  <br>  <br> What other approaches are being used? Curiosity point of view</pre>",
  "messages": [
    {
      "id": "19059",
      "postDate": "01/07/2013 12:43:43",
      "content": "<pre>I have tried following approaches with similar results:<br><br>For all I perform cross-validation on the training data by applying same cutoff times as for PublicLeaderboard.<br><br>a) Extract most recent arrival time for runway/gate from flighthistoryevents.csv file and use them:<br>  Gives same score as benchmark<br>  <br>b) Try to predict the arrival time for runway/gate using the following as predictors: expected runway/gate arrival times, scheduled air time, type of events, .....<br>  SLightly better than benchmark<br>  <br>c) Forget the mins since midnight and instead predict flight times<br>  Same as b)<br>  <br>  <br> What other approaches are being used? Curiosity point of view</pre>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "19060",
      "postDate": "01/07/2013 13:28:35",
      "content": "<p>net-net I have tried to use exp<em>runway</em>arrival<em>mins (mins since midnight) and exp</em>gate<em>arrival</em>mins as the dependent/target variables.</p>\r\n<p>and doing CV on training. </p>\r\n<p>I tried predicting flight time and then arriving at above but it gave near similar results - checking what is the right approach for this so that I don't waste too much time if above approach is wrong</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 19060,
      "author_name": "rkirana",
      "author_url": "",
      "post_date": "01/07/2013 13:28:35",
      "content": "<p>net-net I have tried to use exp<em>runway</em>arrival<em>mins (mins since midnight) and exp</em>gate<em>arrival</em>mins as the dependent/target variables.</p>\r\n<p>and doing CV on training. </p>\r\n<p>I tried predicting flight time and then arriving at above but it gave near similar results - checking what is the right approach for this so that I don't waste too much time if above approach is wrong</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "19059": "",
    "19060": ""
  },
  "source": "meta"
}