{
  "id": 20886,
  "title": "So completely confused about check-in vs (x,y) vs place_id",
  "url": "/competitions/facebook-v-predicting-check-ins/discussion/20886",
  "author_name": "",
  "post_date": "2016-05-12T05:48:54.513Z",
  "votes": 5,
  "comment_count": 8,
  "views": 2047,
  "content": "<p>At first, I thought I got it but after reading the forums, I am confused. </p>\n\n<p>Please take a look at the picture:\n<img src=\"https://drive.google.com/file/d/0B8QmAW82Te3xRlBzMGJHcF9DT0h3WE1MNE10TFRabmRLTmYw/view?usp=sharing\" alt=\"sample map\" title></p>\n\n<p>Do I understand correctly that</p>\n\n<ol>\n<li><p>(x, y) in the dataset correspond to (x1, y1), (x2, y2) coordinates of the users in the picture;</p></li>\n<li><p>place_id corresponds to some assigned id of the CoffeeshopA in the picture?</p></li>\n</ol>",
  "messages": [
    {
      "id": "119667",
      "postDate": "05/12/2016 05:48:54",
      "content": "<p>At first, I thought I got it but after reading the forums, I am confused. </p>\n\n<p>Please take a look at the picture:\n<img src=\"https://drive.google.com/file/d/0B8QmAW82Te3xRlBzMGJHcF9DT0h3WE1MNE10TFRabmRLTmYw/view?usp=sharing\" alt=\"sample map\" title></p>\n\n<p>Do I understand correctly that</p>\n\n<ol>\n<li><p>(x, y) in the dataset correspond to (x1, y1), (x2, y2) coordinates of the users in the picture;</p></li>\n<li><p>place_id corresponds to some assigned id of the CoffeeshopA in the picture?</p></li>\n</ol>",
      "rawMarkdown": "At first, I thought I got it but after reading the forums, I am confused. \r\n\r\nPlease take a look at the picture:\r\n![sample map][1]\r\n\r\nDo I understand correctly that\r\n\r\n1. (x, y) in the dataset correspond to (x1, y1), (x2, y2) coordinates of the users in the picture;\r\n\r\n2. place_id corresponds to some assigned id of the CoffeeshopA in the picture?\r\n\r\n\r\n  [1]: https://drive.google.com/file/d/0B8QmAW82Te3xRlBzMGJHcF9DT0h3WE1MNE10TFRabmRLTmYw/view?usp=sharing",
      "votes": null
    },
    {
      "id": "119719",
      "postDate": "05/12/2016 13:30:06",
      "content": "<p>That is correct understanding. The first training sample with <code>row_id = 0</code>:</p>\n\n<p><code>0,0.7941,9.0809,54,470702,8523065625</code></p>\n\n<p>can be interpreted as:</p>\n\n<p>Somebody at coordinates <code>x = 0.7941</code>, <code>y = 9.0809</code> interacted with place <code>8523065625</code> (checked in). This happened at time <code>470702</code> and the accuracy of this information, whatever that means, is <code>54</code>. We provide no information about the places themselves, i.e., the only insight you can get is through these user interactions.</p>\n\n<p>The objective of this competition is to utilize the knowledge of a large number of past interactions to predict what place a person wants to interact with in the future.</p>",
      "rawMarkdown": "That is correct understanding. The first training sample with `row_id = 0`:\r\n\r\n`0,0.7941,9.0809,54,470702,8523065625`\r\n\r\ncan be interpreted as:\r\n\r\nSomebody at coordinates `x = 0.7941`, `y = 9.0809` interacted with place `8523065625` (checked in). This happened at time `470702` and the accuracy of this information, whatever that means, is `54`. We provide no information about the places themselves, i.e., the only insight you can get is through these user interactions.\r\n\r\nThe objective of this competition is to utilize the knowledge of a large number of past interactions to predict what place a person wants to interact with in the future.",
      "votes": null
    },
    {
      "id": "119737",
      "postDate": "05/12/2016 14:37:01",
      "content": "<p>So there is no guarantee that (x,y) coordinates will be somewhere near a particular place_id, right? (x,y) can be miles away from the actual place, correct?</p>",
      "rawMarkdown": "So there is no guarantee that (x,y) coordinates will be somewhere near a particular place_id, right? (x,y) can be miles away from the actual place, correct?",
      "votes": null
    },
    {
      "id": "119943",
      "postDate": "05/13/2016 21:12:55",
      "content": "<p>I would naturally think that if some person &quot;1&quot; with coordinates (x1, y1) checked into place A, then (x1,y1) should be at or someplace near A.</p>\n\n<p>Pic1: Here, I have sorted the training data by place_id. You can see that  people/events who checked in the same place have similar x &amp; y  coordinate </p>\n\n<p>Pic2: If you plot x against y for a particular place, you get a cluster/ball, indicating that different people/event checked into same place have similar x,y coordinates.</p>",
      "rawMarkdown": "I would naturally think that if some person \"1\" with coordinates (x1, y1) checked into place A, then (x1,y1) should be at or someplace near A.\r\n\r\nPic1: Here, I have sorted the training data by place_id. You can see that  people/events who checked in the same place have similar x & y  coordinate \r\n\r\nPic2: If you plot x against y for a particular place, you get a cluster/ball, indicating that different people/event checked into same place have similar x,y coordinates.",
      "votes": null
    },
    {
      "id": "119994",
      "postDate": "05/14/2016 13:33:03",
      "content": "<p>Thanks for the insight, @vazeer. However, they did mention that a person might have checked in somewhere while physically being at home, but I guess that would be kind of &quot;outliers&quot; so maybe we should ignore them.</p>",
      "rawMarkdown": "Thanks for the insight, @vazeer. However, they did mention that a person might have checked in somewhere while physically being at home, but I guess that would be kind of \"outliers\" so maybe we should ignore them.",
      "votes": null
    },
    {
      "id": "120400",
      "postDate": "05/18/2016 00:37:46",
      "content": "<p>@vazeer I don't see any correlation between the coordinates with place_id. Perhaps the data contains remote checkins? </p>\n\n<pre><code>&gt; print(sprintf(&quot;Correlation of place_id and x: %s&quot;, cor(train$place_id, train$x)))\n[1] &quot;Correlation of place_id and x: -0.00221835109310421&quot;\n&gt; print(sprintf(&quot;Correlation of place_id and y: %s&quot;, cor(train$place_id, train$y)))\n[1] &quot;Correlation of place_id and y: -0.00503800143934364&quot;\n</code></pre>",
      "rawMarkdown": "vazeer I don't see any correlation between the coordinates with place_id. Perhaps the data contains remote checkins? \r\n\r\n    > print(sprintf(\"Correlation of place_id and x: %s\", cor(train$place_id, train$x)))\r\n    [1] \"Correlation of place_id and x: -0.00221835109310421\"\r\n    > print(sprintf(\"Correlation of place_id and y: %s\", cor(train$place_id, train$y)))\r\n    [1] \"Correlation of place_id and y: -0.00503800143934364\"",
      "votes": null
    },
    {
      "id": "120410",
      "postDate": "05/18/2016 01:51:10",
      "content": "<p>@SanjeevBayArea its because your code is trying to find correlation on a class/factor variable. </p>\n\n<p>You need to use ICC (Inter-class correlation coefficient) to find correlation between a class/factor and numerical variable. </p>\n\n<p>Attached picture shows ICC on train data sample.</p>",
      "rawMarkdown": "SanjeevBayArea its because your code is trying to find correlation on a class/factor variable. \r\n\r\nYou need to use ICC (Inter-class correlation coefficient) to find correlation between a class/factor and numerical variable. \r\n\r\nAttached picture shows ICC on train data sample.",
      "votes": null
    },
    {
      "id": "120533",
      "postDate": "05/18/2016 22:13:40",
      "content": "<p>Thanks @vazeer! That was helpful.</p>",
      "rawMarkdown": "Thanks @vazeer! That was helpful.",
      "votes": null
    },
    {
      "id": "120551",
      "postDate": "05/19/2016 00:39:54",
      "content": "<p>Thanks for the confirmation and insights! Now it makes easier to go ahead with Modeling process.</p>",
      "rawMarkdown": "Thanks for the confirmation and insights! Now it makes easier to go ahead with Modeling process.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 119719,
      "author_name": "anaconda",
      "author_url": "",
      "post_date": "05/12/2016 13:30:06",
      "content": "<p>That is correct understanding. The first training sample with <code>row_id = 0</code>:</p>\n\n<p><code>0,0.7941,9.0809,54,470702,8523065625</code></p>\n\n<p>can be interpreted as:</p>\n\n<p>Somebody at coordinates <code>x = 0.7941</code>, <code>y = 9.0809</code> interacted with place <code>8523065625</code> (checked in). This happened at time <code>470702</code> and the accuracy of this information, whatever that means, is <code>54</code>. We provide no information about the places themselves, i.e., the only insight you can get is through these user interactions.</p>\n\n<p>The objective of this competition is to utilize the knowledge of a large number of past interactions to predict what place a person wants to interact with in the future.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 119737,
      "author_name": "osopova",
      "author_url": "",
      "post_date": "05/12/2016 14:37:01",
      "content": "<p>So there is no guarantee that (x,y) coordinates will be somewhere near a particular place_id, right? (x,y) can be miles away from the actual place, correct?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 119943,
      "author_name": "vazeer",
      "author_url": "",
      "post_date": "05/13/2016 21:12:55",
      "content": "<p>I would naturally think that if some person &quot;1&quot; with coordinates (x1, y1) checked into place A, then (x1,y1) should be at or someplace near A.</p>\n\n<p>Pic1: Here, I have sorted the training data by place_id. You can see that  people/events who checked in the same place have similar x &amp; y  coordinate </p>\n\n<p>Pic2: If you plot x against y for a particular place, you get a cluster/ball, indicating that different people/event checked into same place have similar x,y coordinates.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 119994,
      "author_name": "osopova",
      "author_url": "",
      "post_date": "05/14/2016 13:33:03",
      "content": "<p>Thanks for the insight, @vazeer. However, they did mention that a person might have checked in somewhere while physically being at home, but I guess that would be kind of &quot;outliers&quot; so maybe we should ignore them.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 120400,
      "author_name": "gosanjeev",
      "author_url": "",
      "post_date": "05/18/2016 00:37:46",
      "content": "<p>@vazeer I don't see any correlation between the coordinates with place_id. Perhaps the data contains remote checkins? </p>\n\n<pre><code>&gt; print(sprintf(&quot;Correlation of place_id and x: %s&quot;, cor(train$place_id, train$x)))\n[1] &quot;Correlation of place_id and x: -0.00221835109310421&quot;\n&gt; print(sprintf(&quot;Correlation of place_id and y: %s&quot;, cor(train$place_id, train$y)))\n[1] &quot;Correlation of place_id and y: -0.00503800143934364&quot;\n</code></pre>",
      "votes": null,
      "replies": []
    },
    {
      "id": 120410,
      "author_name": "vazeer",
      "author_url": "",
      "post_date": "05/18/2016 01:51:10",
      "content": "<p>@SanjeevBayArea its because your code is trying to find correlation on a class/factor variable. </p>\n\n<p>You need to use ICC (Inter-class correlation coefficient) to find correlation between a class/factor and numerical variable. </p>\n\n<p>Attached picture shows ICC on train data sample.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 120533,
      "author_name": "gosanjeev",
      "author_url": "",
      "post_date": "05/18/2016 22:13:40",
      "content": "<p>Thanks @vazeer! That was helpful.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 120551,
      "author_name": "pavithrarao",
      "author_url": "",
      "post_date": "05/19/2016 00:39:54",
      "content": "<p>Thanks for the confirmation and insights! Now it makes easier to go ahead with Modeling process.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "119667": "At first, I thought I got it but after reading the forums, I am confused. \r\n\r\nPlease take a look at the picture:\r\n![sample map][1]\r\n\r\nDo I understand correctly that\r\n\r\n1. (x, y) in the dataset correspond to (x1, y1), (x2, y2) coordinates of the users in the picture;\r\n\r\n2. place_id corresponds to some assigned id of the CoffeeshopA in the picture?\r\n\r\n\r\n  [1]: https://drive.google.com/file/d/0B8QmAW82Te3xRlBzMGJHcF9DT0h3WE1MNE10TFRabmRLTmYw/view?usp=sharing",
    "119719": "That is correct understanding. The first training sample with `row_id = 0`:\r\n\r\n`0,0.7941,9.0809,54,470702,8523065625`\r\n\r\ncan be interpreted as:\r\n\r\nSomebody at coordinates `x = 0.7941`, `y = 9.0809` interacted with place `8523065625` (checked in). This happened at time `470702` and the accuracy of this information, whatever that means, is `54`. We provide no information about the places themselves, i.e., the only insight you can get is through these user interactions.\r\n\r\nThe objective of this competition is to utilize the knowledge of a large number of past interactions to predict what place a person wants to interact with in the future.",
    "119737": "So there is no guarantee that (x,y) coordinates will be somewhere near a particular place_id, right? (x,y) can be miles away from the actual place, correct?",
    "119943": "I would naturally think that if some person \"1\" with coordinates (x1, y1) checked into place A, then (x1,y1) should be at or someplace near A.\r\n\r\nPic1: Here, I have sorted the training data by place_id. You can see that  people/events who checked in the same place have similar x & y  coordinate \r\n\r\nPic2: If you plot x against y for a particular place, you get a cluster/ball, indicating that different people/event checked into same place have similar x,y coordinates.",
    "119994": "Thanks for the insight, @vazeer. However, they did mention that a person might have checked in somewhere while physically being at home, but I guess that would be kind of \"outliers\" so maybe we should ignore them.",
    "120400": "vazeer I don't see any correlation between the coordinates with place_id. Perhaps the data contains remote checkins? \r\n\r\n    > print(sprintf(\"Correlation of place_id and x: %s\", cor(train$place_id, train$x)))\r\n    [1] \"Correlation of place_id and x: -0.00221835109310421\"\r\n    > print(sprintf(\"Correlation of place_id and y: %s\", cor(train$place_id, train$y)))\r\n    [1] \"Correlation of place_id and y: -0.00503800143934364\"",
    "120410": "SanjeevBayArea its because your code is trying to find correlation on a class/factor variable. \r\n\r\nYou need to use ICC (Inter-class correlation coefficient) to find correlation between a class/factor and numerical variable. \r\n\r\nAttached picture shows ICC on train data sample.",
    "120533": "Thanks @vazeer! That was helpful.",
    "120551": "Thanks for the confirmation and insights! Now it makes easier to go ahead with Modeling process."
  },
  "source": "meta"
}