{
  "id": 10781,
  "title": "hour of day / day of week ctr",
  "url": "/competitions/avazu-ctr-prediction/discussion/10781",
  "author_name": "",
  "post_date": "2014-10-29T07:21:04.837Z",
  "votes": 12,
  "comment_count": 6,
  "views": 2640,
  "content": "<p><strong>Ignore time zone</strong>, just stats from &quot;hour&quot; column.&nbsp;</p>\n<p>hour of day neg pos todal ctr<br>0 1798773 193793 1992566 0.09726<br>1 1759721 231674 1991395 0.11634<br>2 1728651 262402 1991053 0.13179<br>3 1699863 290185 1990048 0.14582<br>4 1654363 333623 1987986 0.16782<br>5 1633564 353140 1986704 0.17775<br>6 1619681 365984 1985665 0.18431<br>7 1600562 384909 1985471 0.19386<br>8 1597868 386428 1984296 0.19474<br>9 1609750 374505 1984255 0.18874<br>10 1596480 388051 1984531 0.19554<br>11 1582541 401835 1984376 0.2025<br>12 1534902 448767 1983669 0.22623<br>13 1512588 470455 1983043 0.23724<br>14 1490031 491718 1981749 0.24812<br>15 1487648 492873 1980521 0.24886<br>16 1526381 455294 1981675 0.22975<br>17 1557604 426540 1984144 0.21497<br>18 1625830 360063 1985893 0.18131<br>19 1702435 286429 1988864 0.14402<br>20 1755768 234614 1990382 0.11787<br>21 1783890 208283 1992173 0.10455<br>22 1795188 197810 1992998 0.09925<br>23 1803590 189304 1992894 0.09499</p>\n<p><img src=\"http://www.kaggle.com/blobs/download/forum-message-attachment-files/1707/hod_ctr.png\" alt=\"ctr per each hour\" width=\"559\" height=\"290\"></p>\n<p>day of week neg pos total ctr<br>1 4023858 746004 4769862 0.1564<br>2 3945278 823939 4769217 0.17276<br>3 7649254 1878871 9528125 0.19719<br>4 7763942 1768174 9532116 0.1855<br>5 7938585 1595557 9534142 0.16735<br>6 4070863 705628 4776491 0.14773<br>7 4065892 710506 4776398 0.14875</p>\n<p><img src=\"http://www.kaggle.com/blobs/download/forum-message-attachment-files/1708/dow_ctr.png\" alt=\"day of week ctr\" width=\"483\" height=\"291\"></p>\n<p>Please note, day of week 1 is Monday. (Wed, Thu, Fri has 2 days of data)</p>",
  "messages": [
    {
      "id": "56962",
      "postDate": "10/29/2014 07:21:04",
      "content": "<p><strong>Ignore time zone</strong>, just stats from &quot;hour&quot; column.&nbsp;</p>\n<p>hour of day neg pos todal ctr<br>0 1798773 193793 1992566 0.09726<br>1 1759721 231674 1991395 0.11634<br>2 1728651 262402 1991053 0.13179<br>3 1699863 290185 1990048 0.14582<br>4 1654363 333623 1987986 0.16782<br>5 1633564 353140 1986704 0.17775<br>6 1619681 365984 1985665 0.18431<br>7 1600562 384909 1985471 0.19386<br>8 1597868 386428 1984296 0.19474<br>9 1609750 374505 1984255 0.18874<br>10 1596480 388051 1984531 0.19554<br>11 1582541 401835 1984376 0.2025<br>12 1534902 448767 1983669 0.22623<br>13 1512588 470455 1983043 0.23724<br>14 1490031 491718 1981749 0.24812<br>15 1487648 492873 1980521 0.24886<br>16 1526381 455294 1981675 0.22975<br>17 1557604 426540 1984144 0.21497<br>18 1625830 360063 1985893 0.18131<br>19 1702435 286429 1988864 0.14402<br>20 1755768 234614 1990382 0.11787<br>21 1783890 208283 1992173 0.10455<br>22 1795188 197810 1992998 0.09925<br>23 1803590 189304 1992894 0.09499</p>\n<p><img src=\"http://www.kaggle.com/blobs/download/forum-message-attachment-files/1707/hod_ctr.png\" alt=\"ctr per each hour\" width=\"559\" height=\"290\"></p>\n<p>day of week neg pos total ctr<br>1 4023858 746004 4769862 0.1564<br>2 3945278 823939 4769217 0.17276<br>3 7649254 1878871 9528125 0.19719<br>4 7763942 1768174 9532116 0.1855<br>5 7938585 1595557 9534142 0.16735<br>6 4070863 705628 4776491 0.14773<br>7 4065892 710506 4776398 0.14875</p>\n<p><img src=\"http://www.kaggle.com/blobs/download/forum-message-attachment-files/1708/dow_ctr.png\" alt=\"day of week ctr\" width=\"483\" height=\"291\"></p>\n<p>Please note, day of week 1 is Monday. (Wed, Thu, Fri has 2 days of data)</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "56971",
      "postDate": "10/29/2014 09:55:32",
      "content": "<p>Interesting.&nbsp;And here is the combination of both plots (based on a sample of training data). By the way, are there any explanations why CTR is maximal on Wednesday?</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "57086",
      "postDate": "10/31/2014 02:33:32",
      "content": "<p>I just noticed that all the test data seems to come from the same day of the week. So if there's this strong of an effect with regards to the day of the week in the training data, it seems like the mean CTR of the test set should be significantly different than the mean CTR of the training set.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "57093",
      "postDate": "10/31/2014 06:57:03",
      "content": "<p>[quote=Miroslav Sabo;56971]</p>\n<p>...&nbsp;By the way, are there any explanations why CTR is maximal on Wednesday?</p>\n<p>[/quote]</p>\n<p>See the Q&amp;A thread:&nbsp;<a href=\"https://www.kaggle.com/c/avazu-ctr-prediction/forums/t/10782/q-a\">https://www.kaggle.com/c/avazu-ctr-prediction/forums/t/10782/q-a</a></p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "57095",
      "postDate": "10/31/2014 07:09:39",
      "content": "<p>[quote=sobrosen;57093]</p>\n<p>[quote=Miroslav Sabo;56971]</p>\n<p>...&nbsp;By the way, are there any explanations why CTR is maximal on Wednesday?</p>\n<p>[/quote]</p>\n<p>See the Q&amp;A thread:&nbsp;<a href=\"https://www.kaggle.com/c/avazu-ctr-prediction/forums/t/10782/q-a\">https://www.kaggle.com/c/avazu-ctr-prediction/forums/t/10782/q-a</a></p>\n<p>[/quote]</p>\n<p>From Q&amp;A 2, does it mean for some reasons there were&nbsp;more click instances on Wed?&nbsp;</p>\n<p>Reasons can be more impressions on Wed or higher ctr on Wed or both.&nbsp;</p>\n<p>In real world, I guess both.&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "57147",
      "postDate": "11/01/2014 01:57:22",
      "content": "<p>[quote=Nicholas Guttenberg;57086]</p>\n<p>I just noticed that all the test data seems to come from the same day of the week. So if there's this strong of an effect with regards to the day of the week in the training data, it seems like the mean CTR of the test set should be significantly different than the mean CTR of the training set.</p>\n<p>[/quote]</p>\n<p>All of the test data seems to come from the same day. Oct 11, 2014. Here is a distribution of observation counts by hour:</p>\n<p>14101100 199449<br>14101101 198809<br>14101102 198948<br>14101103 198879<br>14101104 198854<br>14101105 198626<br>14101106 198922<br>14101107 198728<br>14101108 198588<br>14101109 198153<br>14101110 198244<br>14101111 198488<br>14101112 198680<br>14101113 198534<br>14101114 198677<br>14101115 198658<br>14101116 198530<br>14101117 198749<br>14101118 198365<br>14101119 198436<br>14101120 198692<br>14101121 198997<br>14101122 199114<br>14101123 199281</p>\n<p>Is this intentional? Did I read the data correctly?</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "57148",
      "postDate": "11/01/2014 02:08:29",
      "content": "<p>Never mind. It says in the data description that the test data is for one day.</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 56971,
      "author_name": "miroslavsabo",
      "author_url": "",
      "post_date": "10/29/2014 09:55:32",
      "content": "<p>Interesting.&nbsp;And here is the combination of both plots (based on a sample of training data). By the way, are there any explanations why CTR is maximal on Wednesday?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 57086,
      "author_name": "ngutten",
      "author_url": "",
      "post_date": "10/31/2014 02:33:32",
      "content": "<p>I just noticed that all the test data seems to come from the same day of the week. So if there's this strong of an effect with regards to the day of the week in the training data, it seems like the mean CTR of the test set should be significantly different than the mean CTR of the training set.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 57093,
      "author_name": "sobrosen",
      "author_url": "",
      "post_date": "10/31/2014 06:57:03",
      "content": "<p>[quote=Miroslav Sabo;56971]</p>\n<p>...&nbsp;By the way, are there any explanations why CTR is maximal on Wednesday?</p>\n<p>[/quote]</p>\n<p>See the Q&amp;A thread:&nbsp;<a href=\"https://www.kaggle.com/c/avazu-ctr-prediction/forums/t/10782/q-a\">https://www.kaggle.com/c/avazu-ctr-prediction/forums/t/10782/q-a</a></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 57095,
      "author_name": "soundwaveli00",
      "author_url": "",
      "post_date": "10/31/2014 07:09:39",
      "content": "<p>[quote=sobrosen;57093]</p>\n<p>[quote=Miroslav Sabo;56971]</p>\n<p>...&nbsp;By the way, are there any explanations why CTR is maximal on Wednesday?</p>\n<p>[/quote]</p>\n<p>See the Q&amp;A thread:&nbsp;<a href=\"https://www.kaggle.com/c/avazu-ctr-prediction/forums/t/10782/q-a\">https://www.kaggle.com/c/avazu-ctr-prediction/forums/t/10782/q-a</a></p>\n<p>[/quote]</p>\n<p>From Q&amp;A 2, does it mean for some reasons there were&nbsp;more click instances on Wed?&nbsp;</p>\n<p>Reasons can be more impressions on Wed or higher ctr on Wed or both.&nbsp;</p>\n<p>In real world, I guess both.&nbsp;</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 57147,
      "author_name": "alkhwarizmi",
      "author_url": "",
      "post_date": "11/01/2014 01:57:22",
      "content": "<p>[quote=Nicholas Guttenberg;57086]</p>\n<p>I just noticed that all the test data seems to come from the same day of the week. So if there's this strong of an effect with regards to the day of the week in the training data, it seems like the mean CTR of the test set should be significantly different than the mean CTR of the training set.</p>\n<p>[/quote]</p>\n<p>All of the test data seems to come from the same day. Oct 11, 2014. Here is a distribution of observation counts by hour:</p>\n<p>14101100 199449<br>14101101 198809<br>14101102 198948<br>14101103 198879<br>14101104 198854<br>14101105 198626<br>14101106 198922<br>14101107 198728<br>14101108 198588<br>14101109 198153<br>14101110 198244<br>14101111 198488<br>14101112 198680<br>14101113 198534<br>14101114 198677<br>14101115 198658<br>14101116 198530<br>14101117 198749<br>14101118 198365<br>14101119 198436<br>14101120 198692<br>14101121 198997<br>14101122 199114<br>14101123 199281</p>\n<p>Is this intentional? Did I read the data correctly?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 57148,
      "author_name": "alkhwarizmi",
      "author_url": "",
      "post_date": "11/01/2014 02:08:29",
      "content": "<p>Never mind. It says in the data description that the test data is for one day.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "56962": "",
    "56971": "",
    "57086": "",
    "57093": "",
    "57095": "",
    "57147": "",
    "57148": ""
  },
  "source": "meta"
}