{
  "id": 98670,
  "title": "Significance of 'Additional Data' given only  for train data",
  "url": "/competitions/champs-scalar-coupling/discussion/98670",
  "author_name": "",
  "post_date": "2019-07-05T11:06:32.088996800Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p>There are 5 such  'Additional Data' datasets like potential_energy_df, mulliken_charges_df, scalar_coupling_contributions_df, magnetic_shielding_tensors_df, dipole_moments_df given along with train and test datasets. My doubt is how to use data of these datasets since these are only given for train dataset. Is it like first we have to predict these additional data for test dataset and then use them along with core test data for final prediction.</p>",
  "messages": [
    {
      "id": "568739",
      "postDate": "07/05/2019 11:06:32",
      "content": "<p>There are 5 such  'Additional Data' datasets like potential_energy_df, mulliken_charges_df, scalar_coupling_contributions_df, magnetic_shielding_tensors_df, dipole_moments_df given along with train and test datasets. My doubt is how to use data of these datasets since these are only given for train dataset. Is it like first we have to predict these additional data for test dataset and then use them along with core test data for final prediction.</p>",
      "rawMarkdown": "There are 5 such  'Additional Data' datasets like potential_energy_df, mulliken_charges_df, scalar_coupling_contributions_df, magnetic_shielding_tensors_df, dipole_moments_df given along with train and test datasets. My doubt is how to use data of these datasets since these are only given for train dataset. Is it like first we have to predict these additional data for test dataset and then use them along with core test data for final prediction.",
      "votes": null
    },
    {
      "id": "568740",
      "postDate": "07/05/2019 11:08:06",
      "content": "<p>Did you try searching for kernels about this?\nI have written such kernel long ago: <a href=\"https://www.kaggle.com/artgor/using-meta-features-to-improve-model\">https://www.kaggle.com/artgor/using-meta-features-to-improve-model</a></p>",
      "rawMarkdown": "Did you try searching for kernels about this?\nI have written such kernel long ago: https://www.kaggle.com/artgor/using-meta-features-to-improve-model",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 568740,
      "author_name": "artgor",
      "author_url": "",
      "post_date": "07/05/2019 11:08:06",
      "content": "<p>Did you try searching for kernels about this?\nI have written such kernel long ago: <a href=\"https://www.kaggle.com/artgor/using-meta-features-to-improve-model\">https://www.kaggle.com/artgor/using-meta-features-to-improve-model</a></p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "568739": "There are 5 such  'Additional Data' datasets like potential_energy_df, mulliken_charges_df, scalar_coupling_contributions_df, magnetic_shielding_tensors_df, dipole_moments_df given along with train and test datasets. My doubt is how to use data of these datasets since these are only given for train dataset. Is it like first we have to predict these additional data for test dataset and then use them along with core test data for final prediction.",
    "568740": "Did you try searching for kernels about this?\nI have written such kernel long ago: https://www.kaggle.com/artgor/using-meta-features-to-improve-model"
  },
  "source": "meta"
}