{
  "id": 368373,
  "title": "Some Interesting Times Series on Products",
  "url": "/competitions/otto-recommender-system/discussion/368373",
  "author_name": "",
  "post_date": "2022-11-24T22:25:22.619044800Z",
  "votes": 13,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hi,</p>\n<p>In this <a href=\"https://www.kaggle.com/adaubas/otto-interesting-times-series-eda-on-products\" target=\"_blank\">notebook</a>, there are time series EDA for products which have the most clicks / carts / orders per week.<br>\nYou will see that many products doesn't exhibit regular pattern.</p>\n<p>A product with a lot of clicks / carts / orders one week or one day, will not have necessary a lot of clicks / carts / orders next days. So there should be many useless products to make robust predictions : we should build models and make predictions only for products which have some regular patterns, otherwise we will overfit. </p>\n<p>Do you agree ?</p>\n<p>UPDATE : but unsual clicks/…  on a special product one day could explain clicks/…  on an other product later.</p>",
  "messages": [
    {
      "id": "2042697",
      "postDate": "11/24/2022 22:25:22",
      "content": "<p>Hi,</p>\n<p>In this <a href=\"https://www.kaggle.com/adaubas/otto-interesting-times-series-eda-on-products\" target=\"_blank\">notebook</a>, there are time series EDA for products which have the most clicks / carts / orders per week.<br>\nYou will see that many products doesn't exhibit regular pattern.</p>\n<p>A product with a lot of clicks / carts / orders one week or one day, will not have necessary a lot of clicks / carts / orders next days. So there should be many useless products to make robust predictions : we should build models and make predictions only for products which have some regular patterns, otherwise we will overfit. </p>\n<p>Do you agree ?</p>\n<p>UPDATE : but unsual clicks/…  on a special product one day could explain clicks/…  on an other product later.</p>",
      "rawMarkdown": "Hi,\n\nIn this [notebook](https://www.kaggle.com/adaubas/otto-interesting-times-series-eda-on-products), there are time series EDA for products which have the most clicks / carts / orders per week.\nYou will see that many products doesn't exhibit regular pattern.\n\nA product with a lot of clicks / carts / orders one week or one day, will not have necessary a lot of clicks / carts / orders next days. So there should be many useless products to make robust predictions : we should build models and make predictions only for products which have some regular patterns, otherwise we will overfit. \n\nDo you agree ?\n\nUPDATE : but unsual clicks/...  on a special product one day could explain clicks/...  on an other product later.",
      "votes": null
    },
    {
      "id": "2057269",
      "postDate": "12/06/2022 23:13:33",
      "content": "<p>This EDA is interesting. Thank you for sharing</p>\n<p>I suspect there are occasional promotions and sales which can make items more desirable on certain days.</p>",
      "rawMarkdown": "This EDA is interesting. Thank you for sharing\n\nI suspect there are occasional promotions and sales which can make items more desirable on certain days.",
      "votes": null
    },
    {
      "id": "2067166",
      "postDate": "12/16/2022 13:28:36",
      "content": "<p>I found an heuristic to find promoted items:<br>\n<a href=\"https://www.kaggle.com/code/simonveitner/eda-find-promoted-products?scriptVersionId=113992794\" target=\"_blank\">Here</a> the corresponding notebook.</p>",
      "rawMarkdown": "I found an heuristic to find promoted items:\n[Here](https://www.kaggle.com/code/simonveitner/eda-find-promoted-products?scriptVersionId=113992794) the corresponding notebook.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2057269,
      "author_name": "cdeotte",
      "author_url": "",
      "post_date": "12/06/2022 23:13:33",
      "content": "<p>This EDA is interesting. Thank you for sharing</p>\n<p>I suspect there are occasional promotions and sales which can make items more desirable on certain days.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 2067166,
      "author_name": "simonveitner",
      "author_url": "",
      "post_date": "12/16/2022 13:28:36",
      "content": "<p>I found an heuristic to find promoted items:<br>\n<a href=\"https://www.kaggle.com/code/simonveitner/eda-find-promoted-products?scriptVersionId=113992794\" target=\"_blank\">Here</a> the corresponding notebook.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2042697": "Hi,\n\nIn this [notebook](https://www.kaggle.com/adaubas/otto-interesting-times-series-eda-on-products), there are time series EDA for products which have the most clicks / carts / orders per week.\nYou will see that many products doesn't exhibit regular pattern.\n\nA product with a lot of clicks / carts / orders one week or one day, will not have necessary a lot of clicks / carts / orders next days. So there should be many useless products to make robust predictions : we should build models and make predictions only for products which have some regular patterns, otherwise we will overfit. \n\nDo you agree ?\n\nUPDATE : but unsual clicks/...  on a special product one day could explain clicks/...  on an other product later.",
    "2057269": "This EDA is interesting. Thank you for sharing\n\nI suspect there are occasional promotions and sales which can make items more desirable on certain days.",
    "2067166": "I found an heuristic to find promoted items:\n[Here](https://www.kaggle.com/code/simonveitner/eda-find-promoted-products?scriptVersionId=113992794) the corresponding notebook."
  },
  "source": "meta"
}