{
  "id": 21402,
  "title": "Newbie question - MAP@3 of partial submission",
  "url": "/competitions/facebook-v-predicting-check-ins/discussion/21402",
  "author_name": "",
  "post_date": "2016-06-03T04:52:16.433Z",
  "votes": null,
  "comment_count": 1,
  "views": 465,
  "content": "<p>Is there a way to get feedback on a partial submission of data? I'm rather new to Kaggle and was thinking that it would be very convenient to get MAP@3 for a partial submission of say, half a million rows instead of the full 8 million + rows of the test csv. Thanks in advance!</p>",
  "messages": [
    {
      "id": "122333",
      "postDate": "06/03/2016 04:52:16",
      "content": "<p>Is there a way to get feedback on a partial submission of data? I'm rather new to Kaggle and was thinking that it would be very convenient to get MAP@3 for a partial submission of say, half a million rows instead of the full 8 million + rows of the test csv. Thanks in advance!</p>",
      "rawMarkdown": "Is there a way to get feedback on a partial submission of data? I'm rather new to Kaggle and was thinking that it would be very convenient to get MAP@3 for a partial submission of say, half a million rows instead of the full 8 million + rows of the test csv. Thanks in advance!",
      "votes": null
    },
    {
      "id": "122336",
      "postDate": "06/03/2016 05:11:20",
      "content": "<p>[quote=RamirezAmayaS;122333]</p>\n\n<p>Is there a way to get feedback on a partial submission of data? I'm rather new to Kaggle and was thinking that it would be very convenient to get MAP@3 for a partial submission of say, half a million rows instead of the full 8 million + rows of the test csv. Thanks in advance!</p>\n\n<p>[/quote]\nYou may put 1 M  randomly selected rows with your prediction, and left the other 7 M with only the row_id,\nby the definition of MAP <a href=\"https://www.kaggle.com/wiki/MeanAveragePrecision\">https://www.kaggle.com/wiki/MeanAveragePrecision</a>, for the rest of 7M\nyou get all average precision as 0, so you MAP on the 1M rows is about 8 time the public lead board\nresult on this submission. </p>\n\n<p>You need to make the 1M selection random as the public lead board only select 50% of the data.</p>\n\n<p>Any way, I'll prefer to use split the train data by time to do validation, see the excellent MSB:\n<a href=\"https://www.kaggle.com/zfturbo/facebook-v-predicting-check-ins/msb-with-validation\">https://www.kaggle.com/zfturbo/facebook-v-predicting-check-ins/msb-with-validation</a></p>",
      "rawMarkdown": "[quote=RamirezAmayaS;122333]\r\n\r\nIs there a way to get feedback on a partial submission of data? I'm rather new to Kaggle and was thinking that it would be very convenient to get MAP@3 for a partial submission of say, half a million rows instead of the full 8 million + rows of the test csv. Thanks in advance!\r\n\r\n[/quote]\r\nYou may put 1 M  randomly selected rows with your prediction, and left the other 7 M with only the row_id,\r\nby the definition of MAP https://www.kaggle.com/wiki/MeanAveragePrecision, for the rest of 7M\r\nyou get all average precision as 0, so you MAP on the 1M rows is about 8 time the public lead board\r\nresult on this submission. \r\n\r\nYou need to make the 1M selection random as the public lead board only select 50% of the data.\r\n\r\nAny way, I'll prefer to use split the train data by time to do validation, see the excellent MSB:\r\nhttps://www.kaggle.com/zfturbo/facebook-v-predicting-check-ins/msb-with-validation",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 122336,
      "author_name": "weizhezhang",
      "author_url": "",
      "post_date": "06/03/2016 05:11:20",
      "content": "<p>[quote=RamirezAmayaS;122333]</p>\n\n<p>Is there a way to get feedback on a partial submission of data? I'm rather new to Kaggle and was thinking that it would be very convenient to get MAP@3 for a partial submission of say, half a million rows instead of the full 8 million + rows of the test csv. Thanks in advance!</p>\n\n<p>[/quote]\nYou may put 1 M  randomly selected rows with your prediction, and left the other 7 M with only the row_id,\nby the definition of MAP <a href=\"https://www.kaggle.com/wiki/MeanAveragePrecision\">https://www.kaggle.com/wiki/MeanAveragePrecision</a>, for the rest of 7M\nyou get all average precision as 0, so you MAP on the 1M rows is about 8 time the public lead board\nresult on this submission. </p>\n\n<p>You need to make the 1M selection random as the public lead board only select 50% of the data.</p>\n\n<p>Any way, I'll prefer to use split the train data by time to do validation, see the excellent MSB:\n<a href=\"https://www.kaggle.com/zfturbo/facebook-v-predicting-check-ins/msb-with-validation\">https://www.kaggle.com/zfturbo/facebook-v-predicting-check-ins/msb-with-validation</a></p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "122333": "Is there a way to get feedback on a partial submission of data? I'm rather new to Kaggle and was thinking that it would be very convenient to get MAP@3 for a partial submission of say, half a million rows instead of the full 8 million + rows of the test csv. Thanks in advance!",
    "122336": "[quote=RamirezAmayaS;122333]\r\n\r\nIs there a way to get feedback on a partial submission of data? I'm rather new to Kaggle and was thinking that it would be very convenient to get MAP@3 for a partial submission of say, half a million rows instead of the full 8 million + rows of the test csv. Thanks in advance!\r\n\r\n[/quote]\r\nYou may put 1 M  randomly selected rows with your prediction, and left the other 7 M with only the row_id,\r\nby the definition of MAP https://www.kaggle.com/wiki/MeanAveragePrecision, for the rest of 7M\r\nyou get all average precision as 0, so you MAP on the 1M rows is about 8 time the public lead board\r\nresult on this submission. \r\n\r\nYou need to make the 1M selection random as the public lead board only select 50% of the data.\r\n\r\nAny way, I'll prefer to use split the train data by time to do validation, see the excellent MSB:\r\nhttps://www.kaggle.com/zfturbo/facebook-v-predicting-check-ins/msb-with-validation"
  },
  "source": "meta"
}