{
  "id": 42826,
  "title": "Found an Interesting kernal similar to this competition!!!",
  "url": "/competitions/favorita-grocery-sales-forecasting/discussion/42826",
  "author_name": "",
  "post_date": "2017-11-05T17:58:09.650561Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Below the link to the kernal, it well documented and explained.. all the best.\n<a href=\"https://www.kaggle.com/elenapetrova/time-series-analysis-and-forecasts-with-prophet\">https://www.kaggle.com/elenapetrova/time-series-analysis-and-forecasts-with-prophet</a></p>",
  "messages": [
    {
      "id": "240147",
      "postDate": "11/05/2017 17:58:09",
      "content": "<p>Below the link to the kernal, it well documented and explained.. all the best.\n<a href=\"https://www.kaggle.com/elenapetrova/time-series-analysis-and-forecasts-with-prophet\">https://www.kaggle.com/elenapetrova/time-series-analysis-and-forecasts-with-prophet</a></p>",
      "rawMarkdown": "Below the link to the kernal, it well documented and explained.. all the best.\nhttps://www.kaggle.com/elenapetrova/time-series-analysis-and-forecasts-with-prophet",
      "votes": null
    },
    {
      "id": "241662",
      "postDate": "11/09/2017 14:22:04",
      "content": "<p>I'm currently working on this kind of model. That's really interesting but:</p>\n\n<ul>\n<li>it doesn't work with the item/store combos that don't appear in train set (approx 40k out of 210k)</li>\n<li>it doesn't work well with sparse history</li>\n<li>it consumes a lot of resources (more than 6h of computation on a 32CPU AWS instance)</li>\n<li>I've added some regressors (oil, ...) but there are still unexplained behaviors  </li>\n</ul>\n\n<p>It's a work in progress but the first results are disappointing.</p>",
      "rawMarkdown": "I'm currently working on this kind of model. That's really interesting but:\n\n - it doesn't work with the item/store combos that don't appear in train set (approx 40k out of 210k)\n - it doesn't work well with sparse history\n - it consumes a lot of resources (more than 6h of computation on a 32CPU AWS instance)\n - I've added some regressors (oil, ...) but there are still unexplained behaviors  \n\nIt's a work in progress but the first results are disappointing.",
      "votes": null
    },
    {
      "id": "241719",
      "postDate": "11/09/2017 17:23:35",
      "content": "<p>Sad to hear that it didn't help you much, but thanks for sharing your points which might be helpful for other.</p>",
      "rawMarkdown": "Sad to hear that it didn't help you much, but thanks for sharing your points which might be helpful for other.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 241662,
      "author_name": "ericdfr",
      "author_url": "",
      "post_date": "11/09/2017 14:22:04",
      "content": "<p>I'm currently working on this kind of model. That's really interesting but:</p>\n\n<ul>\n<li>it doesn't work with the item/store combos that don't appear in train set (approx 40k out of 210k)</li>\n<li>it doesn't work well with sparse history</li>\n<li>it consumes a lot of resources (more than 6h of computation on a 32CPU AWS instance)</li>\n<li>I've added some regressors (oil, ...) but there are still unexplained behaviors  </li>\n</ul>\n\n<p>It's a work in progress but the first results are disappointing.</p>",
      "votes": null,
      "replies": [
        {
          "id": 241719,
          "author_name": "yatishbn",
          "author_url": "",
          "post_date": "11/09/2017 17:23:35",
          "content": "<p>Sad to hear that it didn't help you much, but thanks for sharing your points which might be helpful for other.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "240147": "Below the link to the kernal, it well documented and explained.. all the best.\nhttps://www.kaggle.com/elenapetrova/time-series-analysis-and-forecasts-with-prophet",
    "241662": "I'm currently working on this kind of model. That's really interesting but:\n\n - it doesn't work with the item/store combos that don't appear in train set (approx 40k out of 210k)\n - it doesn't work well with sparse history\n - it consumes a lot of resources (more than 6h of computation on a 32CPU AWS instance)\n - I've added some regressors (oil, ...) but there are still unexplained behaviors  \n\nIt's a work in progress but the first results are disappointing.",
    "241719": "Sad to hear that it didn't help you much, but thanks for sharing your points which might be helpful for other."
  },
  "source": "meta"
}