{
  "id": 26884,
  "title": "Pointwise, Pairwise, Listwise...",
  "url": "/competitions/outbrain-click-prediction/discussion/26884",
  "author_name": "",
  "post_date": "2016-12-24T02:21:04.007Z",
  "votes": null,
  "comment_count": 1,
  "views": 311,
  "content": "<p>Is anybody taking an approach other than pointwise? \nI thought about taking a pairwise approach, however, I can't figure out how I would go about doing so given categorical data.</p>",
  "messages": [
    {
      "id": "152205",
      "postDate": "12/24/2016 02:21:04",
      "content": "<p>Is anybody taking an approach other than pointwise? \nI thought about taking a pairwise approach, however, I can't figure out how I would go about doing so given categorical data.</p>",
      "rawMarkdown": "Is anybody taking an approach other than pointwise? \r\nI thought about taking a pairwise approach, however, I can't figure out how I would go about doing so given categorical data.",
      "votes": null
    },
    {
      "id": "153124",
      "postDate": "12/29/2016 22:57:29",
      "content": "<p>I tried doing pairwise comparisons, basically evaluating a linear classifier for each pair plus the context, and used AHP to rank the display. Really had trouble getting it to work better than (or even as well as) ordering by page click probability, though.</p>",
      "rawMarkdown": "I tried doing pairwise comparisons, basically evaluating a linear classifier for each pair plus the context, and used AHP to rank the display. Really had trouble getting it to work better than (or even as well as) ordering by page click probability, though.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 153124,
      "author_name": "midikiman",
      "author_url": "",
      "post_date": "12/29/2016 22:57:29",
      "content": "<p>I tried doing pairwise comparisons, basically evaluating a linear classifier for each pair plus the context, and used AHP to rank the display. Really had trouble getting it to work better than (or even as well as) ordering by page click probability, though.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "152205": "Is anybody taking an approach other than pointwise? \r\nI thought about taking a pairwise approach, however, I can't figure out how I would go about doing so given categorical data.",
    "153124": "I tried doing pairwise comparisons, basically evaluating a linear classifier for each pair plus the context, and used AHP to rank the display. Really had trouble getting it to work better than (or even as well as) ordering by page click probability, though."
  },
  "source": "meta"
}