{
  "id": 11099,
  "title": "Doubts in Variables",
  "url": "/competitions/avazu-ctr-prediction/discussion/11099",
  "author_name": "",
  "post_date": "2014-12-01T03:29:22.287Z",
  "votes": null,
  "comment_count": 6,
  "views": 1913,
  "content": "<p>Hi friends,&nbsp;</p>\n<p>I am new to this type of projects.I have some queries regarding the&nbsp;variables like:</p>\n<p>C1 -- anonymized categorical variable,</p>\n<p>C14-C21 -- anonymized categorical variables.</p>\n\n<p>Do I need to use these variables,if so what variables are these?</p>\n<p>Could you please clarify this query?</p>\n<p>Thanks in advance</p>\n<p>Sandeep.</p>",
  "messages": [
    {
      "id": "59300",
      "postDate": "12/01/2014 03:29:22",
      "content": "<p>Hi friends,&nbsp;</p>\n<p>I am new to this type of projects.I have some queries regarding the&nbsp;variables like:</p>\n<p>C1 -- anonymized categorical variable,</p>\n<p>C14-C21 -- anonymized categorical variables.</p>\n\n<p>Do I need to use these variables,if so what variables are these?</p>\n<p>Could you please clarify this query?</p>\n<p>Thanks in advance</p>\n<p>Sandeep.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "59301",
      "postDate": "12/01/2014 03:39:24",
      "content": "<p>I think in general you would want to use those variables.</p>\n<p>Here is a&nbsp;<a href=\"http://www.kaggle.com/c/tradeshift-text-classification/forums/t/10561/help-hash-values\">forum thread</a> from a recent competition that explains how you can use such variables with <a href=\"http://scikit-learn.org/stable/modules/preprocessing.html#encoding-categorical-features\">one-hot encoding</a>.</p>\n<p>Also, you can check out&nbsp;<a href=\"http://www.kaggle.com/c/avazu-ctr-prediction/forums/t/10927/beat-the-benchmark-with-less-than-1mb-of-memory\">this benckmark</a> which uses the <a href=\"http://en.wikipedia.org/wiki/Feature_hashing\">hash trick</a> to deal with the variables.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "59302",
      "postDate": "12/01/2014 04:11:50",
      "content": "<p>It has also been noted on the forums that:</p>\n<ul>\n<li>C15 and C16 are the width and height of the ad</li>\n<li>C21 ranges from 0 to 255,&nbsp;maybe it's a 8-bit color? or an ascii character?</li>\n</ul>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "59304",
      "postDate": "12/01/2014 05:22:49",
      "content": "<p>Hi All,</p>\n<p>Thank you for your replies.</p>\n<p>Can you please tell me whether do I need to include these variables in the analysis or not?</p>\n<p>I don't think so.What do you say?</p>\n\n<p>Thanks,<br>Sandeep</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "59305",
      "postDate": "12/01/2014 05:31:30",
      "content": "<p>You can leave them out if you want to, but as far as I can tell most of them are helpful. For example C15 and C16 are the dimensions of the ad. Banners and interstitials don't have the same behavior at all, so these features are very valuable...</p>\n\n<p>Why don't you test your model with/without each of these features, and also try some combinations to get a better idea of how they affect your score?</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "59306",
      "postDate": "12/01/2014 05:44:36",
      "content": "<p>For better prediction, these variables are helpful. Check out the relative importance and use them.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "59307",
      "postDate": "12/01/2014 05:45:14",
      "content": "<p>Hi Alexis,</p>\n<p>Thanks for&nbsp;your reply.</p>\n<p>C15 and C16 are fine.I would like to know about the other variables like:</p>\n<p>C16 to C21 .....Are these useful? Because there is no description for these variables.</p>\n<p>Thanks,</p>\n<p>Sandeep</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 59301,
      "author_name": "ggglhf",
      "author_url": "",
      "post_date": "12/01/2014 03:39:24",
      "content": "<p>I think in general you would want to use those variables.</p>\n<p>Here is a&nbsp;<a href=\"http://www.kaggle.com/c/tradeshift-text-classification/forums/t/10561/help-hash-values\">forum thread</a> from a recent competition that explains how you can use such variables with <a href=\"http://scikit-learn.org/stable/modules/preprocessing.html#encoding-categorical-features\">one-hot encoding</a>.</p>\n<p>Also, you can check out&nbsp;<a href=\"http://www.kaggle.com/c/avazu-ctr-prediction/forums/t/10927/beat-the-benchmark-with-less-than-1mb-of-memory\">this benckmark</a> which uses the <a href=\"http://en.wikipedia.org/wiki/Feature_hashing\">hash trick</a> to deal with the variables.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 59302,
      "author_name": "alexistaugeron",
      "author_url": "",
      "post_date": "12/01/2014 04:11:50",
      "content": "<p>It has also been noted on the forums that:</p>\n<ul>\n<li>C15 and C16 are the width and height of the ad</li>\n<li>C21 ranges from 0 to 255,&nbsp;maybe it's a 8-bit color? or an ascii character?</li>\n</ul>",
      "votes": null,
      "replies": []
    },
    {
      "id": 59304,
      "author_name": "sandeepkumar",
      "author_url": "",
      "post_date": "12/01/2014 05:22:49",
      "content": "<p>Hi All,</p>\n<p>Thank you for your replies.</p>\n<p>Can you please tell me whether do I need to include these variables in the analysis or not?</p>\n<p>I don't think so.What do you say?</p>\n\n<p>Thanks,<br>Sandeep</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 59305,
      "author_name": "alexistaugeron",
      "author_url": "",
      "post_date": "12/01/2014 05:31:30",
      "content": "<p>You can leave them out if you want to, but as far as I can tell most of them are helpful. For example C15 and C16 are the dimensions of the ad. Banners and interstitials don't have the same behavior at all, so these features are very valuable...</p>\n\n<p>Why don't you test your model with/without each of these features, and also try some combinations to get a better idea of how they affect your score?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 59306,
      "author_name": "",
      "author_url": "",
      "post_date": "12/01/2014 05:44:36",
      "content": "<p>For better prediction, these variables are helpful. Check out the relative importance and use them.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 59307,
      "author_name": "sandeepkumar",
      "author_url": "",
      "post_date": "12/01/2014 05:45:14",
      "content": "<p>Hi Alexis,</p>\n<p>Thanks for&nbsp;your reply.</p>\n<p>C15 and C16 are fine.I would like to know about the other variables like:</p>\n<p>C16 to C21 .....Are these useful? Because there is no description for these variables.</p>\n<p>Thanks,</p>\n<p>Sandeep</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "59300": "",
    "59301": "",
    "59302": "",
    "59304": "",
    "59305": "",
    "59306": "",
    "59307": ""
  },
  "source": "meta"
}