{
  "id": 66353,
  "title": "Question about the test dataset and the metric you choose to optimize the model",
  "url": "/competitions/ga-customer-revenue-prediction/discussion/66353",
  "author_name": "",
  "post_date": "2018-09-20T22:01:21.022166Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I am fresh here and just start this competition. I noticed that if I use the test dataset to make the prediction, my submission file has some problems with more records than the sample_submission file. If I use the fullVisitorId in sample_submission file to choose the id in test dataset, I will miss some id. And these id are in the training dataset. Which means, we only can give these id 0 for the prediction. And I wrong? Or anyone can help me understand it? I also suffer the problem on choosing sum(log) or log(sum). I am not sure whether any one have better idea about which measurement I should use?  I made a simple test about these two measurement. When I use the sum(log), the cv and lb are very similar. They are 1.8. However, when I use the log(sum), the cv is 1.6 and the pb is 1.9. Is anyone here have the same problem?</p>\n\n<p>Thank you. </p>",
  "messages": [
    {
      "id": "390816",
      "postDate": "09/20/2018 22:01:21",
      "content": "<p>I am fresh here and just start this competition. I noticed that if I use the test dataset to make the prediction, my submission file has some problems with more records than the sample_submission file. If I use the fullVisitorId in sample_submission file to choose the id in test dataset, I will miss some id. And these id are in the training dataset. Which means, we only can give these id 0 for the prediction. And I wrong? Or anyone can help me understand it? I also suffer the problem on choosing sum(log) or log(sum). I am not sure whether any one have better idea about which measurement I should use?  I made a simple test about these two measurement. When I use the sum(log), the cv and lb are very similar. They are 1.8. However, when I use the log(sum), the cv is 1.6 and the pb is 1.9. Is anyone here have the same problem?</p>\n\n<p>Thank you. </p>",
      "rawMarkdown": "I am fresh here and just start this competition. I noticed that if I use the test dataset to make the prediction, my submission file has some problems with more records than the sample_submission file. If I use the fullVisitorId in sample_submission file to choose the id in test dataset, I will miss some id. And these id are in the training dataset. Which means, we only can give these id 0 for the prediction. And I wrong? Or anyone can help me understand it? I also suffer the problem on choosing sum(log) or log(sum). I am not sure whether any one have better idea about which measurement I should use?  I made a simple test about these two measurement. When I use the sum(log), the cv and lb are very similar. They are 1.8. However, when I use the log(sum), the cv is 1.6 and the pb is 1.9. Is anyone here have the same problem?\n\nThank you.",
      "votes": null
    },
    {
      "id": "392932",
      "postDate": "09/24/2018 15:44:53",
      "content": "<p>Note that same visitorIDs in test set come more than once, that's the reason of you having more records in the submission file. Sum up by visitorID ;-)</p>",
      "rawMarkdown": "Note that same visitorIDs in test set come more than once, that's the reason of you having more records in the submission file. Sum up by visitorID ;-)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 392932,
      "author_name": "fingoldo",
      "author_url": "",
      "post_date": "09/24/2018 15:44:53",
      "content": "<p>Note that same visitorIDs in test set come more than once, that's the reason of you having more records in the submission file. Sum up by visitorID ;-)</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "390816": "I am fresh here and just start this competition. I noticed that if I use the test dataset to make the prediction, my submission file has some problems with more records than the sample_submission file. If I use the fullVisitorId in sample_submission file to choose the id in test dataset, I will miss some id. And these id are in the training dataset. Which means, we only can give these id 0 for the prediction. And I wrong? Or anyone can help me understand it? I also suffer the problem on choosing sum(log) or log(sum). I am not sure whether any one have better idea about which measurement I should use?  I made a simple test about these two measurement. When I use the sum(log), the cv and lb are very similar. They are 1.8. However, when I use the log(sum), the cv is 1.6 and the pb is 1.9. Is anyone here have the same problem?\n\nThank you.",
    "392932": "Note that same visitorIDs in test set come more than once, that's the reason of you having more records in the submission file. Sum up by visitorID ;-)"
  },
  "source": "meta"
}