{
  "id": 12380,
  "title": "How to verify result locally??",
  "url": "/competitions/avazu-ctr-prediction/discussion/12380",
  "author_name": "",
  "post_date": "2015-02-02T13:38:05.943Z",
  "votes": null,
  "comment_count": 1,
  "views": 746,
  "content": "<p>Hi, i'm new to kaggle. For this competition, if i generate my output e.g. sampleSubmission.csv, how can i verify the result locally?? It seems on leaderboard, there is a single &quot;0.3818529&quot; score shown?? Do i need to submit first so to check the score?? Thanks.</p>",
  "messages": [
    {
      "id": "63434",
      "postDate": "02/02/2015 13:38:05",
      "content": "<p>Hi, i'm new to kaggle. For this competition, if i generate my output e.g. sampleSubmission.csv, how can i verify the result locally?? It seems on leaderboard, there is a single &quot;0.3818529&quot; score shown?? Do i need to submit first so to check the score?? Thanks.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "63441",
      "postDate": "02/02/2015 16:20:40",
      "content": "<p>The most common way is to use cross validation approaches. See this thread for more information on the debate of which holdout method to use:</p>\n<p>http://www.kaggle.com/c/avazu-ctr-prediction/forums/t/11019/validation-vs-lb</p>\n<p>There are other discussions deeper within the forums as well, but that's a good starting place. Starting late, using the online learning code from tinrtgu is probably a good idea, and it is already set up to run either type of cross-validation:</p>\n<p>http://www.kaggle.com/c/avazu-ctr-prediction/forums/t/10927/beat-the-benchmark-with-less-than-1mb-of-memory</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 63441,
      "author_name": "mlandry",
      "author_url": "",
      "post_date": "02/02/2015 16:20:40",
      "content": "<p>The most common way is to use cross validation approaches. See this thread for more information on the debate of which holdout method to use:</p>\n<p>http://www.kaggle.com/c/avazu-ctr-prediction/forums/t/11019/validation-vs-lb</p>\n<p>There are other discussions deeper within the forums as well, but that's a good starting place. Starting late, using the online learning code from tinrtgu is probably a good idea, and it is already set up to run either type of cross-validation:</p>\n<p>http://www.kaggle.com/c/avazu-ctr-prediction/forums/t/10927/beat-the-benchmark-with-less-than-1mb-of-memory</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "63434": "",
    "63441": ""
  },
  "source": "meta"
}