{
  "id": 21721,
  "title": "Destination latent vectors",
  "url": "/competitions/expedia-hotel-recommendations/discussion/21721",
  "author_name": "CPMP",
  "post_date": "2016-06-16T09:43:07.117000",
  "votes": 0,
  "comment_count": 0,
  "views": 467,
  "content": "<p>Did anyone used these?</p>\n\n<p>I used them to refine predictions for destinations in the test set that did not occur in the training set.  I used k nearest neighbors for that.</p>\n\n<p>I only made predictions for destinations appearing in the training set, then blended nearest neighbor ones to fill remaining destinations.  This improved score a bit.</p>",
  "messages": [
    {
      "id": 124215,
      "postDate": "2016-06-16T09:43:07.117Z",
      "content": "<p>Did anyone used these?</p>\n\n<p>I used them to refine predictions for destinations in the test set that did not occur in the training set.  I used k nearest neighbors for that.</p>\n\n<p>I only made predictions for destinations appearing in the training set, then blended nearest neighbor ones to fill remaining destinations.  This improved score a bit.</p>",
      "rawMarkdown": "Did anyone used these?\r\n\r\n\r\nI used them to refine predictions for destinations in the test set that did not occur in the training set.  I used k nearest neighbors for that.\r\n\r\nI only made predictions for destinations appearing in the training set, then blended nearest neighbor ones to fill remaining destinations.  This improved score a bit."
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "124215": "Did anyone used these?\r\n\r\n\r\nI used them to refine predictions for destinations in the test set that did not occur in the training set.  I used k nearest neighbors for that.\r\n\r\nI only made predictions for destinations appearing in the training set, then blended nearest neighbor ones to fill remaining destinations.  This improved score a bit."
  }
}