{
  "id": 292960,
  "title": "[Updated]Where is the test data?",
  "url": "/competitions/g-research-crypto-forecasting/discussion/292960",
  "author_name": "",
  "post_date": "2021-12-04T01:14:18.402619Z",
  "votes": null,
  "comment_count": 4,
  "views": 0,
  "content": "<p>Today, I joined this competition to get well-understanding ML.<br>\nbut I could not understand what data should be used for testing.<br>\nbecause There is no \"test.csv\" file.</p>\n<p><strong>Could you please let me know below two?</strong></p>\n<p>1) Where is the test data?<br>\n     ==&gt; What data should be used for testing?<br>\n 2) What is \"supplemental_train.csv\" file?<br>\n     ==&gt; I read Data description section, but it is quite difficult for me to understand that description.<br>\n           If possible, Could someone tell me how to get meaning of this supplemental_train.csv?<br>\n           (Simply summarized will be helpful for me…)</p>\n<h2>==below is updated message==</h2>\n<p>I read thorough all overview comments, again. current my understanding is…</p>\n<ul>\n<li>There is no test.csv</li>\n<li>test data will be provided after one month from deadline timing.<br>\n(Because test data will be actual data from market)</li>\n</ul>\n<p>so, What should I do is…</p>\n<ul>\n<li>To make and use both df_train and df_valid which will be divided from train.csv.</li>\n<li>To blushing up the score using these DataFrames provided by train data with tough/robust designed model.<br>\n<strong>Is it correct understanding??</strong></li>\n</ul>",
  "messages": [
    {
      "id": "1605119",
      "postDate": "12/04/2021 01:14:18",
      "content": "<p>Today, I joined this competition to get well-understanding ML.<br>\nbut I could not understand what data should be used for testing.<br>\nbecause There is no \"test.csv\" file.</p>\n<p><strong>Could you please let me know below two?</strong></p>\n<p>1) Where is the test data?<br>\n     ==&gt; What data should be used for testing?<br>\n 2) What is \"supplemental_train.csv\" file?<br>\n     ==&gt; I read Data description section, but it is quite difficult for me to understand that description.<br>\n           If possible, Could someone tell me how to get meaning of this supplemental_train.csv?<br>\n           (Simply summarized will be helpful for me…)</p>\n<h2>==below is updated message==</h2>\n<p>I read thorough all overview comments, again. current my understanding is…</p>\n<ul>\n<li>There is no test.csv</li>\n<li>test data will be provided after one month from deadline timing.<br>\n(Because test data will be actual data from market)</li>\n</ul>\n<p>so, What should I do is…</p>\n<ul>\n<li>To make and use both df_train and df_valid which will be divided from train.csv.</li>\n<li>To blushing up the score using these DataFrames provided by train data with tough/robust designed model.<br>\n<strong>Is it correct understanding??</strong></li>\n</ul>",
      "rawMarkdown": "Today, I joined this competition to get well-understanding ML.\nbut I could not understand what data should be used for testing.\nbecause There is no \"test.csv\" file.\n\n**Could you please let me know below two?**\n\n 1) Where is the test data?\n     ==> What data should be used for testing?\n 2) What is \"supplemental_train.csv\" file?\n     ==> I read Data description section, but it is quite difficult for me to understand that description.\n           If possible, Could someone tell me how to get meaning of this supplemental_train.csv?\n           (Simply summarized will be helpful for me...)\n\n## ==below is updated message==\nI read thorough all overview comments, again. current my understanding is...\n- There is no test.csv\n- test data will be provided after one month from deadline timing.\n  (Because test data will be actual data from market)\n\nso, What should I do is...\n- To make and use both df_train and df_valid which will be divided from train.csv.\n- To blushing up the score using these DataFrames provided by train data with tough/robust designed model.\n**Is it correct understanding??**",
      "votes": null
    },
    {
      "id": "1619077",
      "postDate": "12/15/2021 16:18:46",
      "content": "<p>Hi Jack,</p>\n<p>As I understand it, you have to load the timeseries environment and do a prediction on every batch that iterator gives you. At the moment this iterator just returns some data that is also represented in the tail of the training data set. Only after the competition phase starts, the iterator will be filled with future evaluation data. <br>\nThats also the reason why the current leaderboard is not representative. <br>\nYou have to submit the predictions as follows:</p>\n<p>import gresearch_crypto<br>\nenv = gresearch_crypto.make_env()   # initialize the environment<br>\niter_test = env.iter_test() # an iterator which loops over the test set and sample submission<br>\nfor (test_df, sample_prediction_df) in iter_test:<br>\n    <strong>sample_prediction_df['Target'] = 0  # your prediction goes here!</strong><br>\n    env.predict(sample_prediction_df)   # register your predictions</p>",
      "rawMarkdown": "Hi Jack,\n\nAs I understand it, you have to load the timeseries environment and do a prediction on every batch that iterator gives you. At the moment this iterator just returns some data that is also represented in the tail of the training data set. Only after the competition phase starts, the iterator will be filled with future evaluation data. \nThats also the reason why the current leaderboard is not representative. \nYou have to submit the predictions as follows:\n\nimport gresearch_crypto\nenv = gresearch_crypto.make_env()   # initialize the environment\niter_test = env.iter_test() # an iterator which loops over the test set and sample submission\nfor (test_df, sample_prediction_df) in iter_test:\n    **sample_prediction_df['Target'] = 0  # your prediction goes here!**\n    env.predict(sample_prediction_df)   # register your predictions",
      "votes": null
    },
    {
      "id": "1621852",
      "postDate": "12/18/2021 06:21:25",
      "content": "<p>Hi Wolfgang,</p>\n<p>Thank you for your reply!<br>\nAgain, I'd like to confirm it just in case.</p>\n<p>latest my understanding, This is a code competition.<br>\n==&gt;so, No need to send the result file which will be made from test.csv.</p>\n<p>All I have to do is only to make a code and post it which is learned from train data.<br>\nand this code must be exported such as \"example_sample_submission.csv\" by using \"example_test.csv\" which will be provided after period.</p>\n<p>Is it correct understanding??</p>",
      "rawMarkdown": "Hi Wolfgang,\n\nThank you for your reply!\nAgain, I'd like to confirm it just in case.\n\nlatest my understanding, This is a code competition.\n==>so, No need to send the result file which will be made from test.csv.\n\nAll I have to do is only to make a code and post it which is learned from train data.\nand this code must be exported such as \"example_sample_submission.csv\" by using \"example_test.csv\" which will be provided after period.\n\n Is it correct understanding??",
      "votes": null
    },
    {
      "id": "1621936",
      "postDate": "12/18/2021 07:51:58",
      "content": "<p>Yes, as far as I know you just have to submit the notebook that uses this iterator code part above to predict and call env.predict() <br>\nThe submission file is not really relevant from my perspective as the final competition test will be done on code only.</p>",
      "rawMarkdown": "Yes, as far as I know you just have to submit the notebook that uses this iterator code part above to predict and call env.predict() \nThe submission file is not really relevant from my perspective as the final competition test will be done on code only.",
      "votes": null
    },
    {
      "id": "1623094",
      "postDate": "12/19/2021 14:02:26",
      "content": "<p>OK, Good!<br>\nI fully understood, Thank you:)</p>",
      "rawMarkdown": "OK, Good!\nI fully understood, Thank you:)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1619077,
      "author_name": "wolfgangb33r",
      "author_url": "",
      "post_date": "12/15/2021 16:18:46",
      "content": "<p>Hi Jack,</p>\n<p>As I understand it, you have to load the timeseries environment and do a prediction on every batch that iterator gives you. At the moment this iterator just returns some data that is also represented in the tail of the training data set. Only after the competition phase starts, the iterator will be filled with future evaluation data. <br>\nThats also the reason why the current leaderboard is not representative. <br>\nYou have to submit the predictions as follows:</p>\n<p>import gresearch_crypto<br>\nenv = gresearch_crypto.make_env()   # initialize the environment<br>\niter_test = env.iter_test() # an iterator which loops over the test set and sample submission<br>\nfor (test_df, sample_prediction_df) in iter_test:<br>\n    <strong>sample_prediction_df['Target'] = 0  # your prediction goes here!</strong><br>\n    env.predict(sample_prediction_df)   # register your predictions</p>",
      "votes": null,
      "replies": [
        {
          "id": 1621852,
          "author_name": "jackwilliams3rd",
          "author_url": "",
          "post_date": "12/18/2021 06:21:25",
          "content": "<p>Hi Wolfgang,</p>\n<p>Thank you for your reply!<br>\nAgain, I'd like to confirm it just in case.</p>\n<p>latest my understanding, This is a code competition.<br>\n==&gt;so, No need to send the result file which will be made from test.csv.</p>\n<p>All I have to do is only to make a code and post it which is learned from train data.<br>\nand this code must be exported such as \"example_sample_submission.csv\" by using \"example_test.csv\" which will be provided after period.</p>\n<p>Is it correct understanding??</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1621936,
          "author_name": "wolfgangb33r",
          "author_url": "",
          "post_date": "12/18/2021 07:51:58",
          "content": "<p>Yes, as far as I know you just have to submit the notebook that uses this iterator code part above to predict and call env.predict() <br>\nThe submission file is not really relevant from my perspective as the final competition test will be done on code only.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1623094,
          "author_name": "jackwilliams3rd",
          "author_url": "",
          "post_date": "12/19/2021 14:02:26",
          "content": "<p>OK, Good!<br>\nI fully understood, Thank you:)</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1605119": "Today, I joined this competition to get well-understanding ML.\nbut I could not understand what data should be used for testing.\nbecause There is no \"test.csv\" file.\n\n**Could you please let me know below two?**\n\n 1) Where is the test data?\n     ==> What data should be used for testing?\n 2) What is \"supplemental_train.csv\" file?\n     ==> I read Data description section, but it is quite difficult for me to understand that description.\n           If possible, Could someone tell me how to get meaning of this supplemental_train.csv?\n           (Simply summarized will be helpful for me...)\n\n## ==below is updated message==\nI read thorough all overview comments, again. current my understanding is...\n- There is no test.csv\n- test data will be provided after one month from deadline timing.\n  (Because test data will be actual data from market)\n\nso, What should I do is...\n- To make and use both df_train and df_valid which will be divided from train.csv.\n- To blushing up the score using these DataFrames provided by train data with tough/robust designed model.\n**Is it correct understanding??**",
    "1619077": "Hi Jack,\n\nAs I understand it, you have to load the timeseries environment and do a prediction on every batch that iterator gives you. At the moment this iterator just returns some data that is also represented in the tail of the training data set. Only after the competition phase starts, the iterator will be filled with future evaluation data. \nThats also the reason why the current leaderboard is not representative. \nYou have to submit the predictions as follows:\n\nimport gresearch_crypto\nenv = gresearch_crypto.make_env()   # initialize the environment\niter_test = env.iter_test() # an iterator which loops over the test set and sample submission\nfor (test_df, sample_prediction_df) in iter_test:\n    **sample_prediction_df['Target'] = 0  # your prediction goes here!**\n    env.predict(sample_prediction_df)   # register your predictions",
    "1621852": "Hi Wolfgang,\n\nThank you for your reply!\nAgain, I'd like to confirm it just in case.\n\nlatest my understanding, This is a code competition.\n==>so, No need to send the result file which will be made from test.csv.\n\nAll I have to do is only to make a code and post it which is learned from train data.\nand this code must be exported such as \"example_sample_submission.csv\" by using \"example_test.csv\" which will be provided after period.\n\n Is it correct understanding??",
    "1621936": "Yes, as far as I know you just have to submit the notebook that uses this iterator code part above to predict and call env.predict() \nThe submission file is not really relevant from my perspective as the final competition test will be done on code only.",
    "1623094": "OK, Good!\nI fully understood, Thank you:)"
  },
  "source": "meta"
}