{
  "id": 306669,
  "title": "Do we need to think about \"Recommendetion\"?",
  "url": "/competitions/h-and-m-personalized-fashion-recommendations/discussion/306669",
  "author_name": "",
  "post_date": "2022-02-10T10:53:08.877165200Z",
  "votes": 12,
  "comment_count": 1,
  "views": 0,
  "content": "<p>This competition seems to be held to build the recommendation system in the future.</p>\n<p>However, during the evaluation phase, our prediction results do not interfere with the customer's behavior.</p>\n<p>In a word, should we take this competition as a simple purchase forecast?</p>\n<p>------ postscript -------</p>\n<p>As I read through some papers, I gained a bit of knowledge, which I would like to share with you.</p>\n<p>In this competition we are given the history of each customer's purchase. The fact that these information is given is going to be the difference between \"Recommendation\" and simple purchase prediction.</p>\n<p>To be more precise, it seems that the challenge of how to convert the sparse information associated with each individual into features using some method was mainly done in the area of recommendation tasks.</p>",
  "messages": [
    {
      "id": "1684227",
      "postDate": "02/10/2022 10:53:08",
      "content": "<p>This competition seems to be held to build the recommendation system in the future.</p>\n<p>However, during the evaluation phase, our prediction results do not interfere with the customer's behavior.</p>\n<p>In a word, should we take this competition as a simple purchase forecast?</p>\n<p>------ postscript -------</p>\n<p>As I read through some papers, I gained a bit of knowledge, which I would like to share with you.</p>\n<p>In this competition we are given the history of each customer's purchase. The fact that these information is given is going to be the difference between \"Recommendation\" and simple purchase prediction.</p>\n<p>To be more precise, it seems that the challenge of how to convert the sparse information associated with each individual into features using some method was mainly done in the area of recommendation tasks.</p>",
      "rawMarkdown": "This competition seems to be held to build the recommendation system in the future.\n\nHowever, during the evaluation phase, our prediction results do not interfere with the customer's behavior.\n\nIn a word, should we take this competition as a simple purchase forecast?\n\n\n------ postscript -------\n\nAs I read through some papers, I gained a bit of knowledge, which I would like to share with you.\n\nIn this competition we are given the history of each customer's purchase. The fact that these information is given is going to be the difference between \"Recommendation\" and simple purchase prediction.\n\nTo be more precise, it seems that the challenge of how to convert the sparse information associated with each individual into features using some method was mainly done in the area of recommendation tasks.",
      "votes": null
    },
    {
      "id": "1718899",
      "postDate": "03/11/2022 09:09:24",
      "content": "<p><a href=\"https://www.kaggle.com/loto610\" target=\"_blank\">@loto610</a> </p>\n<p>Yes, it is a purchase prediction. But then that is a way to build offline recommender systems. Basically, you are using historical data to verify that how close you could recommend to a Customer's actual purchase.<br>\nUnless you can deploy you system online and record customer interactions with your recommendations and subsequent purchases, this will be the only way to build a recommender system.<br>\nAlso, as a second step, such a recommender system may be deployed in real life and then fine tuned.</p>",
      "rawMarkdown": "loto610 \n\nYes, it is a purchase prediction. But then that is a way to build offline recommender systems. Basically, you are using historical data to verify that how close you could recommend to a Customer's actual purchase.\nUnless you can deploy you system online and record customer interactions with your recommendations and subsequent purchases, this will be the only way to build a recommender system.\nAlso, as a second step, such a recommender system may be deployed in real life and then fine tuned.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1718899,
      "author_name": "atulverma",
      "author_url": "",
      "post_date": "03/11/2022 09:09:24",
      "content": "<p><a href=\"https://www.kaggle.com/loto610\" target=\"_blank\">@loto610</a> </p>\n<p>Yes, it is a purchase prediction. But then that is a way to build offline recommender systems. Basically, you are using historical data to verify that how close you could recommend to a Customer's actual purchase.<br>\nUnless you can deploy you system online and record customer interactions with your recommendations and subsequent purchases, this will be the only way to build a recommender system.<br>\nAlso, as a second step, such a recommender system may be deployed in real life and then fine tuned.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1684227": "This competition seems to be held to build the recommendation system in the future.\n\nHowever, during the evaluation phase, our prediction results do not interfere with the customer's behavior.\n\nIn a word, should we take this competition as a simple purchase forecast?\n\n\n------ postscript -------\n\nAs I read through some papers, I gained a bit of knowledge, which I would like to share with you.\n\nIn this competition we are given the history of each customer's purchase. The fact that these information is given is going to be the difference between \"Recommendation\" and simple purchase prediction.\n\nTo be more precise, it seems that the challenge of how to convert the sparse information associated with each individual into features using some method was mainly done in the area of recommendation tasks.",
    "1718899": "loto610 \n\nYes, it is a purchase prediction. But then that is a way to build offline recommender systems. Basically, you are using historical data to verify that how close you could recommend to a Customer's actual purchase.\nUnless you can deploy you system online and record customer interactions with your recommendations and subsequent purchases, this will be the only way to build a recommender system.\nAlso, as a second step, such a recommender system may be deployed in real life and then fine tuned."
  },
  "source": "meta"
}