{
  "id": 156998,
  "title": "Don't underestimate the Tabular data!!! [LB: .818]",
  "url": "/competitions/siim-isic-melanoma-classification/discussion/156998",
  "author_name": "",
  "post_date": "2020-06-08T20:27:22.437536300Z",
  "votes": 16,
  "comment_count": 10,
  "views": 0,
  "content": "<p>Though the tabular data seems less important at first, we can do so much with it. With some feature engieering I got <strong>.78</strong> <a href=\"https://www.kaggle.com/awsaf49/xgboost-tabular-data-ml-cv-86-lb-79\">here</a> .Before that <strong>.69</strong> was best score in public notebook. Finally after some effort I was able to get <strong>.818</strong> without using any image, deep learning method or ensemble. This result can be further improved by using ensemble.....</p>",
  "messages": [
    {
      "id": "878772",
      "postDate": "06/08/2020 20:27:22",
      "content": "<p>Though the tabular data seems less important at first, we can do so much with it. With some feature engieering I got <strong>.78</strong> <a href=\"https://www.kaggle.com/awsaf49/xgboost-tabular-data-ml-cv-86-lb-79\">here</a> .Before that <strong>.69</strong> was best score in public notebook. Finally after some effort I was able to get <strong>.818</strong> without using any image, deep learning method or ensemble. This result can be further improved by using ensemble.....</p>",
      "rawMarkdown": "Though the tabular data seems less important at first, we can do so much with it. With some feature engieering I got **.78** [here](https://www.kaggle.com/awsaf49/xgboost-tabular-data-ml-cv-86-lb-79) .Before that **.69** was best score in public notebook. Finally after some effort I was able to get **.818** without using any image, deep learning method or ensemble. This result can be further improved by using ensemble.....",
      "votes": null
    },
    {
      "id": "878867",
      "postDate": "06/09/2020 00:48:57",
      "content": "<p>Great job. Yes, there is much information hiding in the meta data</p>",
      "rawMarkdown": "Great job. Yes, there is much information hiding in the meta data",
      "votes": null
    },
    {
      "id": "878946",
      "postDate": "06/09/2020 03:59:49",
      "content": "<p>Thanks.</p>",
      "rawMarkdown": "Thanks.",
      "votes": null
    },
    {
      "id": "881354",
      "postDate": "06/11/2020 00:56:02",
      "content": "<p>I just public notebook with cv: 0.819 and pl: 0.819 using only 4 features.</p>",
      "rawMarkdown": "I just public notebook with cv: 0.819 and pl: 0.819 using only 4 features.",
      "votes": null
    },
    {
      "id": "881468",
      "postDate": "06/11/2020 04:19:21",
      "content": "<p>great work, which notebook?</p>",
      "rawMarkdown": "great work, which notebook?",
      "votes": null
    },
    {
      "id": "881776",
      "postDate": "06/11/2020 11:05:51",
      "content": "<p>Amazing. নোটবুকটা পড়ে ভালো লেগেছে, সেভ করে রাখলাম 😃 </p>",
      "rawMarkdown": "Amazing. নোটবুকটা পড়ে ভালো লেগেছে, সেভ করে রাখলাম 😃",
      "votes": null
    },
    {
      "id": "881830",
      "postDate": "06/11/2020 11:55:53",
      "content": "<p>It's strange but in my experiments blending \"metadata-submissions\" with \"only-imagedata-submissions\" didn't get any improvements in public score.  May be meta data didn't contains new information. Perhaps all relevant information is contaned in image data. </p>",
      "rawMarkdown": "It's strange but in my experiments blending \"metadata-submissions\" with \"only-imagedata-submissions\" didn't get any improvements in public score.  May be meta data didn't contains new information. Perhaps all relevant information is contaned in image data.",
      "votes": null
    },
    {
      "id": "881864",
      "postDate": "06/11/2020 12:26:01",
      "content": "<p><a href=\"/awsaf49\">@awsaf49</a> I think it's this one. <br>\n<a href=\"https://www.kaggle.com/ragnar123/catboost-metadata\">https://www.kaggle.com/ragnar123/catboost-metadata</a></p>",
      "rawMarkdown": "awsaf49 I think it's this one.  \nhttps://www.kaggle.com/ragnar123/catboost-metadata",
      "votes": null
    },
    {
      "id": "881914",
      "postDate": "06/11/2020 13:24:21",
      "content": "<p>Great work!👍 </p>",
      "rawMarkdown": "Great work!👍",
      "votes": null
    },
    {
      "id": "882039",
      "postDate": "06/11/2020 14:47:14",
      "content": "<p>Great one !! I think you utilized each fold data for prediction. I didn't use that one....</p>",
      "rawMarkdown": "Great one !! I think you utilized each fold data for prediction. I didn't use that one....",
      "votes": null
    },
    {
      "id": "882042",
      "postDate": "06/11/2020 14:48:47",
      "content": "<p>Thanks😄 </p>",
      "rawMarkdown": "Thanks😄",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 878867,
      "author_name": "cdeotte",
      "author_url": "",
      "post_date": "06/09/2020 00:48:57",
      "content": "<p>Great job. Yes, there is much information hiding in the meta data</p>",
      "votes": null,
      "replies": [
        {
          "id": 878946,
          "author_name": "awsaf49",
          "author_url": "",
          "post_date": "06/09/2020 03:59:49",
          "content": "<p>Thanks.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 881354,
      "author_name": "ragnar123",
      "author_url": "",
      "post_date": "06/11/2020 00:56:02",
      "content": "<p>I just public notebook with cv: 0.819 and pl: 0.819 using only 4 features.</p>",
      "votes": null,
      "replies": [
        {
          "id": 881468,
          "author_name": "awsaf49",
          "author_url": "",
          "post_date": "06/11/2020 04:19:21",
          "content": "<p>great work, which notebook?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 881864,
          "author_name": "sdeagggg",
          "author_url": "",
          "post_date": "06/11/2020 12:26:01",
          "content": "<p><a href=\"/awsaf49\">@awsaf49</a> I think it's this one. <br>\n<a href=\"https://www.kaggle.com/ragnar123/catboost-metadata\">https://www.kaggle.com/ragnar123/catboost-metadata</a></p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 882039,
          "author_name": "awsaf49",
          "author_url": "",
          "post_date": "06/11/2020 14:47:14",
          "content": "<p>Great one !! I think you utilized each fold data for prediction. I didn't use that one....</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 881776,
      "author_name": "ipythonx",
      "author_url": "",
      "post_date": "06/11/2020 11:05:51",
      "content": "<p>Amazing. নোটবুকটা পড়ে ভালো লেগেছে, সেভ করে রাখলাম 😃 </p>",
      "votes": null,
      "replies": [
        {
          "id": 882042,
          "author_name": "awsaf49",
          "author_url": "",
          "post_date": "06/11/2020 14:48:47",
          "content": "<p>Thanks😄 </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 881830,
      "author_name": "andreyzotov",
      "author_url": "",
      "post_date": "06/11/2020 11:55:53",
      "content": "<p>It's strange but in my experiments blending \"metadata-submissions\" with \"only-imagedata-submissions\" didn't get any improvements in public score.  May be meta data didn't contains new information. Perhaps all relevant information is contaned in image data. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 881914,
      "author_name": "pratikbarua",
      "author_url": "",
      "post_date": "06/11/2020 13:24:21",
      "content": "<p>Great work!👍 </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "878772": "Though the tabular data seems less important at first, we can do so much with it. With some feature engieering I got **.78** [here](https://www.kaggle.com/awsaf49/xgboost-tabular-data-ml-cv-86-lb-79) .Before that **.69** was best score in public notebook. Finally after some effort I was able to get **.818** without using any image, deep learning method or ensemble. This result can be further improved by using ensemble.....",
    "878867": "Great job. Yes, there is much information hiding in the meta data",
    "878946": "Thanks.",
    "881354": "I just public notebook with cv: 0.819 and pl: 0.819 using only 4 features.",
    "881468": "great work, which notebook?",
    "881776": "Amazing. নোটবুকটা পড়ে ভালো লেগেছে, সেভ করে রাখলাম 😃",
    "881830": "It's strange but in my experiments blending \"metadata-submissions\" with \"only-imagedata-submissions\" didn't get any improvements in public score.  May be meta data didn't contains new information. Perhaps all relevant information is contaned in image data.",
    "881864": "awsaf49 I think it's this one.  \nhttps://www.kaggle.com/ragnar123/catboost-metadata",
    "881914": "Great work!👍",
    "882039": "Great one !! I think you utilized each fold data for prediction. I didn't use that one....",
    "882042": "Thanks😄"
  },
  "source": "meta"
}