{
  "id": 175382,
  "title": "Blend simple tabular model can improve PB score",
  "url": "/competitions/siim-isic-melanoma-classification/discussion/175382",
  "author_name": "",
  "post_date": "2020-08-18T03:56:16.262416900Z",
  "votes": 2,
  "comment_count": 3,
  "views": 0,
  "content": "<p>I don't believe complex tabular model (XGB or ML model),<br>\nI believe target encoding model  ,like <a href=\"https://www.kaggle.com/titericz/simple-baseline\" target=\"_blank\">this</a>.<br>\nIt can blend well with CNN model. </p>\n<p>By the way,<br>\nI think I make a mistake,I used including external data to do tabular model .(It high CV score than  use original data  ).<br>\nIf I used original data  to do tabular model ,I will get 0.9472 PB score.I miss a solo gold medal.</p>",
  "messages": [
    {
      "id": "974793",
      "postDate": "08/18/2020 03:56:16",
      "content": "<p>I don't believe complex tabular model (XGB or ML model),<br>\nI believe target encoding model  ,like <a href=\"https://www.kaggle.com/titericz/simple-baseline\" target=\"_blank\">this</a>.<br>\nIt can blend well with CNN model. </p>\n<p>By the way,<br>\nI think I make a mistake,I used including external data to do tabular model .(It high CV score than  use original data  ).<br>\nIf I used original data  to do tabular model ,I will get 0.9472 PB score.I miss a solo gold medal.</p>",
      "rawMarkdown": "I don't believe complex tabular model (XGB or ML model),\nI believe target encoding model  ,like [this](https://www.kaggle.com/titericz/simple-baseline).\nIt can blend well with CNN model. \n \n\nBy the way,\nI think I make a mistake,I used including external data to do tabular model .(It high CV score than  use original data  ).\nIf I used original data  to do tabular model ,I will get 0.9472 PB score.I miss a solo gold medal.",
      "votes": null
    },
    {
      "id": "974803",
      "postDate": "08/18/2020 04:00:05",
      "content": "<p>Congrats on your great finish. I agree that that simple meta data notebook increases LB a lot. What were your image models like?</p>",
      "rawMarkdown": "Congrats on your great finish. I agree that that simple meta data notebook increases LB a lot. What were your image models like?",
      "votes": null
    },
    {
      "id": "974813",
      "postDate": "08/18/2020 04:06:59",
      "content": "<p>My image model only have 0.9351 PB score.I blend public notebook models with my model.(0.15 my model ,0.85  public notebook models )</p>",
      "rawMarkdown": "My image model only have 0.9351 PB score.I blend public notebook models with my model.(0.15 my model ,0.85  public notebook models )",
      "votes": null
    },
    {
      "id": "974828",
      "postDate": "08/18/2020 04:15:30",
      "content": "<p>Wow, great job Fate. Smartly done.</p>",
      "rawMarkdown": "Wow, great job Fate. Smartly done.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 974803,
      "author_name": "cdeotte",
      "author_url": "",
      "post_date": "08/18/2020 04:00:05",
      "content": "<p>Congrats on your great finish. I agree that that simple meta data notebook increases LB a lot. What were your image models like?</p>",
      "votes": null,
      "replies": [
        {
          "id": 974813,
          "author_name": "chihantsai",
          "author_url": "",
          "post_date": "08/18/2020 04:06:59",
          "content": "<p>My image model only have 0.9351 PB score.I blend public notebook models with my model.(0.15 my model ,0.85  public notebook models )</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 974828,
          "author_name": "cdeotte",
          "author_url": "",
          "post_date": "08/18/2020 04:15:30",
          "content": "<p>Wow, great job Fate. Smartly done.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "974793": "I don't believe complex tabular model (XGB or ML model),\nI believe target encoding model  ,like [this](https://www.kaggle.com/titericz/simple-baseline).\nIt can blend well with CNN model. \n \n\nBy the way,\nI think I make a mistake,I used including external data to do tabular model .(It high CV score than  use original data  ).\nIf I used original data  to do tabular model ,I will get 0.9472 PB score.I miss a solo gold medal.",
    "974803": "Congrats on your great finish. I agree that that simple meta data notebook increases LB a lot. What were your image models like?",
    "974813": "My image model only have 0.9351 PB score.I blend public notebook models with my model.(0.15 my model ,0.85  public notebook models )",
    "974828": "Wow, great job Fate. Smartly done."
  },
  "source": "meta"
}