{
  "id": 56807,
  "title": "how to accelerate rolling with groupby?",
  "url": "/competitions/talkingdata-adtracking-fraud-detection/discussion/56807",
  "author_name": "",
  "post_date": "2018-05-15T08:36:31.060878400Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p>in this competition,  i want to generate features like in [os,ip,app] dimension , last five/ten/30 minites, the click number. But the pandas groupby with rolling is really slow and i failed to generate these features . anyone can share a efficient way for this kind of feature? thanks </p>",
  "messages": [
    {
      "id": "328879",
      "postDate": "05/15/2018 08:36:31",
      "content": "<p>in this competition,  i want to generate features like in [os,ip,app] dimension , last five/ten/30 minites, the click number. But the pandas groupby with rolling is really slow and i failed to generate these features . anyone can share a efficient way for this kind of feature? thanks </p>",
      "rawMarkdown": "in this competition,  i want to generate features like in [os,ip,app] dimension , last five/ten/30 minites, the click number. But the pandas groupby with rolling is really slow and i failed to generate these features . anyone can share a efficient way for this kind of feature? thanks",
      "votes": null
    },
    {
      "id": "328880",
      "postDate": "05/15/2018 08:39:47",
      "content": "<p>in addition ,  there are time index duplicates which are not supported in pandas in dimension like [ip] (because there are a lot clicks happen in one second). this problem also stop me to achieve time rolling features.</p>",
      "rawMarkdown": "in addition ,  there are time index duplicates which are not supported in pandas in dimension like [ip] (because there are a lot clicks happen in one second). this problem also stop me to achieve time rolling features.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 328880,
      "author_name": "",
      "author_url": "",
      "post_date": "05/15/2018 08:39:47",
      "content": "<p>in addition ,  there are time index duplicates which are not supported in pandas in dimension like [ip] (because there are a lot clicks happen in one second). this problem also stop me to achieve time rolling features.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "328879": "in this competition,  i want to generate features like in [os,ip,app] dimension , last five/ten/30 minites, the click number. But the pandas groupby with rolling is really slow and i failed to generate these features . anyone can share a efficient way for this kind of feature? thanks",
    "328880": "in addition ,  there are time index duplicates which are not supported in pandas in dimension like [ip] (because there are a lot clicks happen in one second). this problem also stop me to achieve time rolling features."
  },
  "source": "meta"
}