{
  "id": 20836,
  "title": "this might be useful~",
  "url": "/competitions/expedia-hotel-recommendations/discussion/20836",
  "author_name": "",
  "post_date": "2016-05-10T16:18:18.977Z",
  "votes": 3,
  "comment_count": 4,
  "views": 1388,
  "content": "<p>In case someone wondering training on parts of the data,\nhere is some CV info aiming for some qualitative trend:</p>\n\n<h1>num of row = 360000,  MAP5 = 0.285</h1>\n\n<h1>num of row = 3800000,  MAP5 = 0.383</h1>\n\n<h1>num of row = all,  MAP5 = 0.501</h1>\n\n<p>hope it helps    </p>",
  "messages": [
    {
      "id": "119478",
      "postDate": "05/10/2016 16:18:18",
      "content": "<p>In case someone wondering training on parts of the data,\nhere is some CV info aiming for some qualitative trend:</p>\n\n<h1>num of row = 360000,  MAP5 = 0.285</h1>\n\n<h1>num of row = 3800000,  MAP5 = 0.383</h1>\n\n<h1>num of row = all,  MAP5 = 0.501</h1>\n\n<p>hope it helps    </p>",
      "rawMarkdown": "In case someone wondering training on parts of the data,\r\nhere is some CV info aiming for some qualitative trend:\r\n \r\n# num of row = 360000,  MAP5 = 0.285\r\n# num of row = 3800000,  MAP5 = 0.383\r\n# num of row = all,  MAP5 = 0.501\r\n\r\nhope it helps",
      "votes": null
    },
    {
      "id": "121145",
      "postDate": "05/24/2016 13:16:35",
      "content": "<p>This is useful, but with all data, training the model will be very slow.</p>",
      "rawMarkdown": "This is useful, but with all data, training the model will be very slow.",
      "votes": null
    },
    {
      "id": "121160",
      "postDate": "05/24/2016 16:34:29",
      "content": "<p>What hardware specs you have? I have problem loading all the data in Python on my 24GB ram PC. I noticed R consume less memory!</p>",
      "rawMarkdown": "What hardware specs you have? I have problem loading all the data in Python on my 24GB ram PC. I noticed R consume less memory!",
      "votes": null
    },
    {
      "id": "121384",
      "postDate": "05/26/2016 01:42:37",
      "content": "<p>What kind of method you used to train the model?</p>",
      "rawMarkdown": "What kind of method you used to train the model?",
      "votes": null
    },
    {
      "id": "121397",
      "postDate": "05/26/2016 03:41:06",
      "content": "<p>@YiTang</p>\n\n<p>16GB ram, 4-core i7, specs don't matter too much, it might be a memory management issue lies in panda(python 2.7), changing to short-integer(or boolean) data types helps in this case.</p>\n\n<p>@FengLi</p>\n\n<p>k-way interaction + manual decision tree, group features should also help</p>",
      "rawMarkdown": "YiTang\r\n\r\n16GB ram, 4-core i7, specs don't matter too much, it might be a memory management issue lies in panda(python 2.7), changing to short-integer(or boolean) data types helps in this case.\r\n\r\n@FengLi\r\n\r\nk-way interaction + manual decision tree, group features should also help",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 121145,
      "author_name": "masterliu",
      "author_url": "",
      "post_date": "05/24/2016 13:16:35",
      "content": "<p>This is useful, but with all data, training the model will be very slow.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 121160,
      "author_name": "yimacs",
      "author_url": "",
      "post_date": "05/24/2016 16:34:29",
      "content": "<p>What hardware specs you have? I have problem loading all the data in Python on my 24GB ram PC. I noticed R consume less memory!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 121384,
      "author_name": "beedata",
      "author_url": "",
      "post_date": "05/26/2016 01:42:37",
      "content": "<p>What kind of method you used to train the model?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 121397,
      "author_name": "aviatorx",
      "author_url": "",
      "post_date": "05/26/2016 03:41:06",
      "content": "<p>@YiTang</p>\n\n<p>16GB ram, 4-core i7, specs don't matter too much, it might be a memory management issue lies in panda(python 2.7), changing to short-integer(or boolean) data types helps in this case.</p>\n\n<p>@FengLi</p>\n\n<p>k-way interaction + manual decision tree, group features should also help</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "119478": "In case someone wondering training on parts of the data,\r\nhere is some CV info aiming for some qualitative trend:\r\n \r\n# num of row = 360000,  MAP5 = 0.285\r\n# num of row = 3800000,  MAP5 = 0.383\r\n# num of row = all,  MAP5 = 0.501\r\n\r\nhope it helps",
    "121145": "This is useful, but with all data, training the model will be very slow.",
    "121160": "What hardware specs you have? I have problem loading all the data in Python on my 24GB ram PC. I noticed R consume less memory!",
    "121384": "What kind of method you used to train the model?",
    "121397": "YiTang\r\n\r\n16GB ram, 4-core i7, specs don't matter too much, it might be a memory management issue lies in panda(python 2.7), changing to short-integer(or boolean) data types helps in this case.\r\n\r\n@FengLi\r\n\r\nk-way interaction + manual decision tree, group features should also help"
  },
  "source": "meta"
}