{
  "id": 370404,
  "title": "Minimum /Maximum History data ",
  "url": "/competitions/otto-recommender-system/discussion/370404",
  "author_name": "",
  "post_date": "2022-12-04T09:57:56.115902400Z",
  "votes": 3,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hello, <br>\nI am looking for the answers to following questions.</p>\n<ul>\n<li>Minimum/Maximum span of History data for prediction of \"aid\" of different type.? </li>\n<li>Do we need to predict all aid of any given specific session?</li>\n</ul>\n<p>Thanks</p>",
  "messages": [
    {
      "id": "2054636",
      "postDate": "12/04/2022 09:57:56",
      "content": "<p>Hello, <br>\nI am looking for the answers to following questions.</p>\n<ul>\n<li>Minimum/Maximum span of History data for prediction of \"aid\" of different type.? </li>\n<li>Do we need to predict all aid of any given specific session?</li>\n</ul>\n<p>Thanks</p>",
      "rawMarkdown": "Hello, \nI am looking for the answers to following questions.\n\n- Minimum/Maximum span of History data for prediction of \"aid\" of different type.? \n- Do we need to predict all aid of any given specific session?\n\nThanks",
      "votes": null
    },
    {
      "id": "2055787",
      "postDate": "12/05/2022 12:50:52",
      "content": "<p>Ad. 1 - we are predicting the first followup click and all the carts and orders to come. But, in the test set, given how it was created, the max span of time between the end of any of the sessions and the end of the test set will be equal to 7 days.<br>\nAd. 2 - Yes. Since we don't know up front if the session that was truncated includes clicks/carts/orders, we have to supply all the predictions. But we will only be scored on a given row if there is ground truth for it in the test set.</p>\n<p>If you'd like to study how all this can look in practice, I created a validation dataset from the 4th week of train using a script from the organizer's repo on github. Please find the dataset here: <a href=\"https://www.kaggle.com/datasets/radek1/otto-train-and-test-data-for-local-validation\" target=\"_blank\">OTTO train and validation (extracted from train)</a></p>",
      "rawMarkdown": "Ad. 1 - we are predicting the first followup click and all the carts and orders to come. But, in the test set, given how it was created, the max span of time between the end of any of the sessions and the end of the test set will be equal to 7 days.\nAd. 2 - Yes. Since we don't know up front if the session that was truncated includes clicks/carts/orders, we have to supply all the predictions. But we will only be scored on a given row if there is ground truth for it in the test set.\n\nIf you'd like to study how all this can look in practice, I created a validation dataset from the 4th week of train using a script from the organizer's repo on github. Please find the dataset here: [OTTO train and validation (extracted from train)](https://www.kaggle.com/datasets/radek1/otto-train-and-test-data-for-local-validation)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2055787,
      "author_name": "radek1",
      "author_url": "",
      "post_date": "12/05/2022 12:50:52",
      "content": "<p>Ad. 1 - we are predicting the first followup click and all the carts and orders to come. But, in the test set, given how it was created, the max span of time between the end of any of the sessions and the end of the test set will be equal to 7 days.<br>\nAd. 2 - Yes. Since we don't know up front if the session that was truncated includes clicks/carts/orders, we have to supply all the predictions. But we will only be scored on a given row if there is ground truth for it in the test set.</p>\n<p>If you'd like to study how all this can look in practice, I created a validation dataset from the 4th week of train using a script from the organizer's repo on github. Please find the dataset here: <a href=\"https://www.kaggle.com/datasets/radek1/otto-train-and-test-data-for-local-validation\" target=\"_blank\">OTTO train and validation (extracted from train)</a></p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2054636": "Hello, \nI am looking for the answers to following questions.\n\n- Minimum/Maximum span of History data for prediction of \"aid\" of different type.? \n- Do we need to predict all aid of any given specific session?\n\nThanks",
    "2055787": "Ad. 1 - we are predicting the first followup click and all the carts and orders to come. But, in the test set, given how it was created, the max span of time between the end of any of the sessions and the end of the test set will be equal to 7 days.\nAd. 2 - Yes. Since we don't know up front if the session that was truncated includes clicks/carts/orders, we have to supply all the predictions. But we will only be scored on a given row if there is ground truth for it in the test set.\n\nIf you'd like to study how all this can look in practice, I created a validation dataset from the 4th week of train using a script from the organizer's repo on github. Please find the dataset here: [OTTO train and validation (extracted from train)](https://www.kaggle.com/datasets/radek1/otto-train-and-test-data-for-local-validation)"
  },
  "source": "meta"
}