{
  "id": 541500,
  "title": "Attention:We can get more data.",
  "url": "/competitions/brist1d/discussion/541500",
  "author_name": "",
  "post_date": "2024-10-20T00:03:59.496428400Z",
  "votes": 14,
  "comment_count": 5,
  "views": 0,
  "content": "<p>Although the competition team provided us with a task of predicting the next hour according to 6 hours, we can change the task to, for example, predicting the next hour according to  4 hours. After this conversion, not only does the training data increase, but the testing data can also provide training data without affecting the prediction of the testing data.</p>",
  "messages": [
    {
      "id": "3022784",
      "postDate": "10/20/2024 00:03:59",
      "content": "<p>Although the competition team provided us with a task of predicting the next hour according to 6 hours, we can change the task to, for example, predicting the next hour according to  4 hours. After this conversion, not only does the training data increase, but the testing data can also provide training data without affecting the prediction of the testing data.</p>",
      "rawMarkdown": "Although the competition team provided us with a task of predicting the next hour according to 6 hours, we can change the task to, for example, predicting the next hour according to  4 hours. After this conversion, not only does the training data increase, but the testing data can also provide training data without affecting the prediction of the testing data.",
      "votes": null
    },
    {
      "id": "3026199",
      "postDate": "10/23/2024 14:35:36",
      "content": "<p>Acutually i do not understand it well how can we increase data . Can you tell me in more details </p>",
      "rawMarkdown": "Acutually i do not understand it well how can we increase data . Can you tell me in more details",
      "votes": null
    },
    {
      "id": "3026206",
      "postDate": "10/23/2024 14:40:22",
      "content": "<p>For example, if we have a piece of data with 72 time steps, we can take the 1st to 48th time steps as X, and use the 49th as Y. Then, we can take the 2nd to 49th time steps as X, and use the 50th as Y, and so on.</p>",
      "rawMarkdown": "For example, if we have a piece of data with 72 time steps, we can take the 1st to 48th time steps as X, and use the 49th as Y. Then, we can take the 2nd to 49th time steps as X, and use the 50th as Y, and so on.",
      "votes": null
    },
    {
      "id": "3026216",
      "postDate": "10/23/2024 14:45:20",
      "content": "<p>Oh! I See. Also I have look into your Notebook Also .</p>",
      "rawMarkdown": "Oh! I See. Also I have look into your Notebook Also .",
      "votes": null
    },
    {
      "id": "3026221",
      "postDate": "10/23/2024 14:46:56",
      "content": "<p>Thanks. I will give it a shot.</p>",
      "rawMarkdown": "Thanks. I will give it a shot.",
      "votes": null
    },
    {
      "id": "3036746",
      "postDate": "11/04/2024 22:09:00",
      "content": "<p>That would be interesting to check when reducing the previous data to 4 hours instead of 6 hours, how much does that decrease the accuracy of prediction for the next hour. Also could check whether reduction in performance could be outweighed by having a larger number of training examples. This would lose data from 5 or 6 hours ago, but would still keep the more useful recent data. <br>\nIn fact presumably a similar process has already been applied to this training set by the provider, in order to reduce this data to 6 hour lengths from originally much longer data.</p>",
      "rawMarkdown": "That would be interesting to check when reducing the previous data to 4 hours instead of 6 hours, how much does that decrease the accuracy of prediction for the next hour. Also could check whether reduction in performance could be outweighed by having a larger number of training examples. This would lose data from 5 or 6 hours ago, but would still keep the more useful recent data. \nIn fact presumably a similar process has already been applied to this training set by the provider, in order to reduce this data to 6 hour lengths from originally much longer data.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3026199,
      "author_name": "",
      "author_url": "",
      "post_date": "10/23/2024 14:35:36",
      "content": "<p>Acutually i do not understand it well how can we increase data . Can you tell me in more details </p>",
      "votes": null,
      "replies": [
        {
          "id": 3026206,
          "author_name": "yunsuxiaozi",
          "author_url": "",
          "post_date": "10/23/2024 14:40:22",
          "content": "<p>For example, if we have a piece of data with 72 time steps, we can take the 1st to 48th time steps as X, and use the 49th as Y. Then, we can take the 2nd to 49th time steps as X, and use the 50th as Y, and so on.</p>",
          "votes": null,
          "replies": [
            {
              "id": 3026216,
              "author_name": "",
              "author_url": "",
              "post_date": "10/23/2024 14:45:20",
              "content": "<p>Oh! I See. Also I have look into your Notebook Also .</p>",
              "votes": null,
              "replies": [
                {
                  "id": 3026221,
                  "author_name": "",
                  "author_url": "",
                  "post_date": "10/23/2024 14:46:56",
                  "content": "<p>Thanks. I will give it a shot.</p>",
                  "votes": null,
                  "replies": []
                }
              ]
            }
          ]
        }
      ]
    },
    {
      "id": 3036746,
      "author_name": "neilvaughan",
      "author_url": "",
      "post_date": "11/04/2024 22:09:00",
      "content": "<p>That would be interesting to check when reducing the previous data to 4 hours instead of 6 hours, how much does that decrease the accuracy of prediction for the next hour. Also could check whether reduction in performance could be outweighed by having a larger number of training examples. This would lose data from 5 or 6 hours ago, but would still keep the more useful recent data. <br>\nIn fact presumably a similar process has already been applied to this training set by the provider, in order to reduce this data to 6 hour lengths from originally much longer data.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3022784": "Although the competition team provided us with a task of predicting the next hour according to 6 hours, we can change the task to, for example, predicting the next hour according to  4 hours. After this conversion, not only does the training data increase, but the testing data can also provide training data without affecting the prediction of the testing data.",
    "3026199": "Acutually i do not understand it well how can we increase data . Can you tell me in more details",
    "3026206": "For example, if we have a piece of data with 72 time steps, we can take the 1st to 48th time steps as X, and use the 49th as Y. Then, we can take the 2nd to 49th time steps as X, and use the 50th as Y, and so on.",
    "3026216": "Oh! I See. Also I have look into your Notebook Also .",
    "3026221": "Thanks. I will give it a shot.",
    "3036746": "That would be interesting to check when reducing the previous data to 4 hours instead of 6 hours, how much does that decrease the accuracy of prediction for the next hour. Also could check whether reduction in performance could be outweighed by having a larger number of training examples. This would lose data from 5 or 6 hours ago, but would still keep the more useful recent data. \nIn fact presumably a similar process has already been applied to this training set by the provider, in order to reduce this data to 6 hour lengths from originally much longer data."
  },
  "source": "meta"
}