{
  "id": 20915,
  "title": "Mapping Place ID to the Grid",
  "url": "/competitions/facebook-v-predicting-check-ins/discussion/20915",
  "author_name": "",
  "post_date": "2016-05-12T20:13:12.187000",
  "votes": 0,
  "comment_count": 0,
  "views": 508,
  "content": "<p>So I was thinking a little about this and I realized that we have to create our own target attribute. </p>\n\n<p>We don't know exactly what accuracy means, so we'll have to do some guesswork. For now let's assume that low accuracies are good. To find the coordinate of PLACE_ID we need to subset set the relevant xy coordinates. I was thinking something like:</p>\n\n<p>Sphere(place_id) = {x_1, y_1 | min(acc_{place_id}) }</p>\n\n<p>From here we can identify the center of the circle formed by the min accuracies and actually get the target location. </p>\n\n<p>Edit: I just had the thought that maybe PLACE_ID is the unique identifier that we might be trying to map to and we should use TIME as a way of structuring panel data. </p>",
  "messages": [
    {
      "id": 119795,
      "postDate": "2016-05-12T20:13:12.187Z",
      "content": "<p>So I was thinking a little about this and I realized that we have to create our own target attribute. </p>\n\n<p>We don't know exactly what accuracy means, so we'll have to do some guesswork. For now let's assume that low accuracies are good. To find the coordinate of PLACE_ID we need to subset set the relevant xy coordinates. I was thinking something like:</p>\n\n<p>Sphere(place_id) = {x_1, y_1 | min(acc_{place_id}) }</p>\n\n<p>From here we can identify the center of the circle formed by the min accuracies and actually get the target location. </p>\n\n<p>Edit: I just had the thought that maybe PLACE_ID is the unique identifier that we might be trying to map to and we should use TIME as a way of structuring panel data. </p>",
      "rawMarkdown": "So I was thinking a little about this and I realized that we have to create our own target attribute. \r\n\r\nWe don't know exactly what accuracy means, so we'll have to do some guesswork. For now let's assume that low accuracies are good. To find the coordinate of PLACE_ID we need to subset set the relevant xy coordinates. I was thinking something like:\r\n\r\nSphere(place_id) = {x_1, y_1 | min(acc_{place_id}) }\r\n\r\nFrom here we can identify the center of the circle formed by the min accuracies and actually get the target location. \r\n\r\nEdit: I just had the thought that maybe PLACE_ID is the unique identifier that we might be trying to map to and we should use TIME as a way of structuring panel data. \r\n"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "119795": "So I was thinking a little about this and I realized that we have to create our own target attribute. \r\n\r\nWe don't know exactly what accuracy means, so we'll have to do some guesswork. For now let's assume that low accuracies are good. To find the coordinate of PLACE_ID we need to subset set the relevant xy coordinates. I was thinking something like:\r\n\r\nSphere(place_id) = {x_1, y_1 | min(acc_{place_id}) }\r\n\r\nFrom here we can identify the center of the circle formed by the min accuracies and actually get the target location. \r\n\r\nEdit: I just had the thought that maybe PLACE_ID is the unique identifier that we might be trying to map to and we should use TIME as a way of structuring panel data. \r\n"
  }
}