{
  "id": 434134,
  "title": "How Do You Manage Your Analytical Environment in Kaggle Competitions?",
  "url": "/competitions/cafa-5-protein-function-prediction/discussion/434134",
  "author_name": "",
  "post_date": "2023-08-24T05:24:21.346840Z",
  "votes": 4,
  "comment_count": 1,
  "views": 0,
  "content": "<p>How do you manage your analytical environment in Kaggle competitions? In our team, we've primarily shared datasets and source code within the Kaggle environment. We divided our EDA and feature engineering into small code snippets and ended up creating as many as 1300 pieces of code for this contest 😂. However, by managing the datasets in a sophisticated way and striving to create loosely coupled groups of code, we managed to avoid collapse!</p>\n<p>Additionally, some permanent versions of the code were managed on GitHub. I believe there are various approaches, such as \"one experiment, one script\" or \"one experiment, one config.\" Feel free to share what worked well or didn't work for you!</p>\n<p>Thank you in advance for your insights!</p>",
  "messages": [
    {
      "id": "2405754",
      "postDate": "08/24/2023 05:24:21",
      "content": "<p>How do you manage your analytical environment in Kaggle competitions? In our team, we've primarily shared datasets and source code within the Kaggle environment. We divided our EDA and feature engineering into small code snippets and ended up creating as many as 1300 pieces of code for this contest 😂. However, by managing the datasets in a sophisticated way and striving to create loosely coupled groups of code, we managed to avoid collapse!</p>\n<p>Additionally, some permanent versions of the code were managed on GitHub. I believe there are various approaches, such as \"one experiment, one script\" or \"one experiment, one config.\" Feel free to share what worked well or didn't work for you!</p>\n<p>Thank you in advance for your insights!</p>",
      "rawMarkdown": "How do you manage your analytical environment in Kaggle competitions? In our team, we've primarily shared datasets and source code within the Kaggle environment. We divided our EDA and feature engineering into small code snippets and ended up creating as many as 1300 pieces of code for this contest 😂. However, by managing the datasets in a sophisticated way and striving to create loosely coupled groups of code, we managed to avoid collapse!\n\nAdditionally, some permanent versions of the code were managed on GitHub. I believe there are various approaches, such as \"one experiment, one script\" or \"one experiment, one config.\" Feel free to share what worked well or didn't work for you!\n\nThank you in advance for your insights!",
      "votes": null
    },
    {
      "id": "2405904",
      "postDate": "08/24/2023 06:38:51",
      "content": "<p>I am also very interested in this issue.</p>\n<p>I basically did the entire contest by myself in order to avoid problems in communication as well as code understanding or environment.</p>",
      "rawMarkdown": "I am also very interested in this issue.\n\nI basically did the entire contest by myself in order to avoid problems in communication as well as code understanding or environment.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2405904,
      "author_name": "impimp",
      "author_url": "",
      "post_date": "08/24/2023 06:38:51",
      "content": "<p>I am also very interested in this issue.</p>\n<p>I basically did the entire contest by myself in order to avoid problems in communication as well as code understanding or environment.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2405754": "How do you manage your analytical environment in Kaggle competitions? In our team, we've primarily shared datasets and source code within the Kaggle environment. We divided our EDA and feature engineering into small code snippets and ended up creating as many as 1300 pieces of code for this contest 😂. However, by managing the datasets in a sophisticated way and striving to create loosely coupled groups of code, we managed to avoid collapse!\n\nAdditionally, some permanent versions of the code were managed on GitHub. I believe there are various approaches, such as \"one experiment, one script\" or \"one experiment, one config.\" Feel free to share what worked well or didn't work for you!\n\nThank you in advance for your insights!",
    "2405904": "I am also very interested in this issue.\n\nI basically did the entire contest by myself in order to avoid problems in communication as well as code understanding or environment."
  },
  "source": "meta"
}