{
  "id": 418143,
  "title": "2nd Place Solution",
  "content": "<p>Sorry for late sharing. First of all, thanks to competition organizers for hosting this interesting competition and my great teammates(@hyd,@rib). And congrats my friend rib for becoming GM.</p>\n<p>Our solution is simple, we found information related to \"visit_month\" very useful and information related to proteins are useless. So we focus on the samples, structures and models.</p>\n<h3>Samples</h3>\n<p>We used samples with visit_month in \"[0, 6, 12, 18, 24 , 36, 48, 60, 72, 84]\" for training. And found that our cv is much better correlated to LB.</p>\n<h3>Features</h3>\n<ul>\n<li>All about “visit_month”<br>\n1) use visit_month as meta feature<br>\n2) gap between visit_month and last visit_month<br>\n3) times of visit for each patient</li>\n<li>Little about “protein”<br>\nuse NPX’s ratio of each patient, instead of the original values<br>\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1069377%2F39cefc5f531ed2c7b4d105fe209e44e6%2F1.png?generation=1687139866930379&amp;alt=media\" alt=\"\"></li>\n</ul>\n<h3>Structure</h3>\n<ol>\n<li>16 labels<br>\nPredict each “updrs” for predicted_month_diff=0,6,12,24<br>\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1069377%2Fb41bb7ba7b53f7f259961020b11df086%2F2.png?generation=1687139918111830&amp;alt=media\" alt=\"\"></li>\n<li>4 labels<br>\nUse predicted_month_diff as a feature, and predict each one’s 4 updrs<br>\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1069377%2Fb277c3ac795f150c6f4ae892dc4e485e%2F3.png?generation=1687139946997350&amp;alt=media\" alt=\"\"></li>\n</ol>\n<h3>Models</h3>\n<p>Simple MLP with different structures and parameters.Finally we blends with:</p>\n<ol>\n<li>Models training in different numbers of labels</li>\n<li>Models training in different structure of network</li>\n<li>Models training in different parameters of network</li>\n</ol>\n<p>code:<a href=\"url\" target=\"_blank\">https://www.kaggle.com/code/dc5e964768ef56302a32/2nd-solution</a></p>",
  "comments": [
    {
      "id": 2651010,
      "author_name": "Krish L Sharma",
      "author_url": "",
      "post_date": "2024-02-13T20:15:22.800000",
      "content": "<p>Congratulations on your breakthrough. I would really like to see your code for learning purpose. could you please tell me where to find your code?</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 2308926,
      "author_name": "Majid Ahmad Khan",
      "author_url": "",
      "post_date": "2023-06-19T10:00:38.947000",
      "content": "<p>Congrats on getting 2nd position</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 2308568,
      "author_name": "serangu",
      "author_url": "",
      "post_date": "2023-06-19T04:49:59.950000",
      "content": "<p><a href=\"https://www.kaggle.com/chenxin1991\" target=\"_blank\">@chenxin1991</a> thank you for sharing solution! </p>",
      "votes": 0,
      "replies": []
    }
  ],
  "comment_count": 3,
  "messages": [
    {
      "id": 2308477,
      "postDate": "2023-06-19T01:55:43.587Z",
      "content": "<p>Sorry for late sharing. First of all, thanks to competition organizers for hosting this interesting competition and my great teammates(@hyd,@rib). And congrats my friend rib for becoming GM.</p>\n<p>Our solution is simple, we found information related to \"visit_month\" very useful and information related to proteins are useless. So we focus on the samples, structures and models.</p>\n<h3>Samples</h3>\n<p>We used samples with visit_month in \"[0, 6, 12, 18, 24 , 36, 48, 60, 72, 84]\" for training. And found that our cv is much better correlated to LB.</p>\n<h3>Features</h3>\n<ul>\n<li>All about “visit_month”<br>\n1) use visit_month as meta feature<br>\n2) gap between visit_month and last visit_month<br>\n3) times of visit for each patient</li>\n<li>Little about “protein”<br>\nuse NPX’s ratio of each patient, instead of the original values<br>\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1069377%2F39cefc5f531ed2c7b4d105fe209e44e6%2F1.png?generation=1687139866930379&amp;alt=media\" alt=\"\"></li>\n</ul>\n<h3>Structure</h3>\n<ol>\n<li>16 labels<br>\nPredict each “updrs” for predicted_month_diff=0,6,12,24<br>\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1069377%2Fb41bb7ba7b53f7f259961020b11df086%2F2.png?generation=1687139918111830&amp;alt=media\" alt=\"\"></li>\n<li>4 labels<br>\nUse predicted_month_diff as a feature, and predict each one’s 4 updrs<br>\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1069377%2Fb277c3ac795f150c6f4ae892dc4e485e%2F3.png?generation=1687139946997350&amp;alt=media\" alt=\"\"></li>\n</ol>\n<h3>Models</h3>\n<p>Simple MLP with different structures and parameters.Finally we blends with:</p>\n<ol>\n<li>Models training in different numbers of labels</li>\n<li>Models training in different structure of network</li>\n<li>Models training in different parameters of network</li>\n</ol>\n<p>code:<a href=\"url\" target=\"_blank\">https://www.kaggle.com/code/dc5e964768ef56302a32/2nd-solution</a></p>",
      "rawMarkdown": "Sorry for late sharing. First of all, thanks to competition organizers for hosting this interesting competition and my great teammates(@hyd,@rib). And congrats my friend rib for becoming GM.\n\nOur solution is simple, we found information related to \"visit_month\" very useful and information related to proteins are useless. So we focus on the samples, structures and models.\n\n### Samples\nWe used samples with visit_month in \"[0, 6, 12, 18, 24 , 36, 48, 60, 72, 84]\" for training. And found that our cv is much better correlated to LB.\n\n### Features\n- All about “visit_month”\n1) use visit_month as meta feature\n2) gap between visit_month and last visit_month\n3) times of visit for each patient\n- Little about “protein”\nuse NPX’s ratio of each patient, instead of the original values\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1069377%2F39cefc5f531ed2c7b4d105fe209e44e6%2F1.png?generation=1687139866930379&alt=media)\n### Structure\n1. 16 labels\nPredict each “updrs” for predicted_month_diff=0,6,12,24\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1069377%2Fb41bb7ba7b53f7f259961020b11df086%2F2.png?generation=1687139918111830&alt=media)\n2. 4 labels\nUse predicted_month_diff as a feature, and predict each one’s 4 updrs\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1069377%2Fb277c3ac795f150c6f4ae892dc4e485e%2F3.png?generation=1687139946997350&alt=media)\n### Models\nSimple MLP with different structures and parameters.Finally we blends with:\n1. Models training in different numbers of labels\n2. Models training in different structure of network\n3. Models training in different parameters of network\n\ncode:[https://www.kaggle.com/code/dc5e964768ef56302a32/2nd-solution](url)\n\n\n\n\n\n",
      "votes": 10
    },
    {
      "id": 2651010,
      "postDate": "2024-02-13T20:15:22.800Z",
      "content": "<p>Congratulations on your breakthrough. I would really like to see your code for learning purpose. could you please tell me where to find your code?</p>",
      "rawMarkdown": "Congratulations on your breakthrough. I would really like to see your code for learning purpose. could you please tell me where to find your code?"
    },
    {
      "id": 2308926,
      "postDate": "2023-06-19T10:00:38.947Z",
      "content": "<p>Congrats on getting 2nd position</p>",
      "rawMarkdown": "Congrats on getting 2nd position"
    },
    {
      "id": 2308568,
      "postDate": "2023-06-19T04:49:59.950Z",
      "content": "<p><a href=\"https://www.kaggle.com/chenxin1991\" target=\"_blank\">@chenxin1991</a> thank you for sharing solution! </p>",
      "rawMarkdown": "@chenxin1991 thank you for sharing solution! "
    }
  ],
  "complete": true,
  "source": "live",
  "fetched_at": "2026-10-19T14:11:40Z",
  "competition_id": 37794,
  "competition_slug": "amp-parkinsons-disease-progression-prediction",
  "team_id": 10018342,
  "solution_url": "/competitions/amp-parkinsons-disease-progression-prediction/writeups/no-luck-all-skill-2nd-place-solution"
}