{
  "id": 124167,
  "title": "What do we expect to see with: \"(and interactions!)\"",
  "url": "/competitions/cat-in-the-dat-ii/discussion/124167",
  "author_name": "",
  "post_date": "2020-01-02T11:24:54.505730900Z",
  "votes": 7,
  "comment_count": 6,
  "views": 0,
  "content": "<p>The subtitle of this Competition includes that expression : (and interactions!). Has anyone the idea of what kind of Interactions? How it changes the approaches?</p>",
  "messages": [
    {
      "id": "708518",
      "postDate": "01/02/2020 11:24:54",
      "content": "<p>The subtitle of this Competition includes that expression : (and interactions!). Has anyone the idea of what kind of Interactions? How it changes the approaches?</p>",
      "rawMarkdown": "The subtitle of this Competition includes that expression : (and interactions!). Has anyone the idea of what kind of Interactions? How it changes the approaches?",
      "votes": null
    },
    {
      "id": "713288",
      "postDate": "01/08/2020 05:40:02",
      "content": "<p>Hello,\nIn the first \"Categorical feature encoding challenge\", every feature were independant, there were no interaction between features. In this new challenge, features are not independant : hence tree methods should outperform logisitic regression in this new challenge (logistic regression won the first challenge).</p>",
      "rawMarkdown": "Hello,\nIn the first \"Categorical feature encoding challenge\", every feature were independant, there were no interaction between features. In this new challenge, features are not independant : hence tree methods should outperform logisitic regression in this new challenge (logistic regression won the first challenge).",
      "votes": null
    },
    {
      "id": "713612",
      "postDate": "01/08/2020 13:20:47",
      "content": "<p>Thank you for the answer Aldparis . Now I understood the difference between the two Competitions.</p>",
      "rawMarkdown": "Thank you for the answer Aldparis . Now I understood the difference between the two Competitions.",
      "votes": null
    },
    {
      "id": "713644",
      "postDate": "01/08/2020 13:54:50",
      "content": "<p>J'ai oublié: Mes Félicitations M. Daubas  pour le votre accomplissement dans la dernière Competition (2eme place). Bonne chance.</p>",
      "rawMarkdown": "J'ai oublié: Mes Félicitations M. Daubas  pour le votre accomplissement dans la dernière Competition (2eme place). Bonne chance.",
      "votes": null
    },
    {
      "id": "713914",
      "postDate": "01/08/2020 19:47:54",
      "content": "<p>merci et bonne chance aussi !</p>",
      "rawMarkdown": "merci et bonne chance aussi !",
      "votes": null
    },
    {
      "id": "717714",
      "postDate": "01/13/2020 14:11:56",
      "content": "<p><a href=\"/adaubas\">@adaubas</a> how to check which features are dependent to the others?</p>",
      "rawMarkdown": "adaubas how to check which features are dependent to the others?",
      "votes": null
    },
    {
      "id": "721791",
      "postDate": "01/17/2020 17:40:08",
      "content": "<p><a href=\"/nnikolay\">@nnikolay</a> Get feature correlation.</p>\n\n<p><a href=\"/adaubas\">@adaubas</a>, you say</p>\n\n<blockquote>\n  <p>hence tree methods should outperform logisitic regression in this new challenge</p>\n</blockquote>\n\n<p>If I'm understanding correctly, feature dependence/interaction causes colinearity in logistic regression, making weights unstable. Tree methods like random forest or gradient boosting is not affected by colinearity caused by feature dependence. Is this a correct elaboration of your statement?</p>",
      "rawMarkdown": "nnikolay Get feature correlation.\n\n@adaubas, you say\n\n&gt; hence tree methods should outperform logisitic regression in this new challenge\n\nIf I'm understanding correctly, feature dependence/interaction causes colinearity in logistic regression, making weights unstable. Tree methods like random forest or gradient boosting is not affected by colinearity caused by feature dependence. Is this a correct elaboration of your statement?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 713288,
      "author_name": "adaubas",
      "author_url": "",
      "post_date": "01/08/2020 05:40:02",
      "content": "<p>Hello,\nIn the first \"Categorical feature encoding challenge\", every feature were independant, there were no interaction between features. In this new challenge, features are not independant : hence tree methods should outperform logisitic regression in this new challenge (logistic regression won the first challenge).</p>",
      "votes": null,
      "replies": [
        {
          "id": 713612,
          "author_name": "mpwolke",
          "author_url": "",
          "post_date": "01/08/2020 13:20:47",
          "content": "<p>Thank you for the answer Aldparis . Now I understood the difference between the two Competitions.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 717714,
          "author_name": "nnikolay",
          "author_url": "",
          "post_date": "01/13/2020 14:11:56",
          "content": "<p><a href=\"/adaubas\">@adaubas</a> how to check which features are dependent to the others?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 721791,
          "author_name": "pairwiseseq",
          "author_url": "",
          "post_date": "01/17/2020 17:40:08",
          "content": "<p><a href=\"/nnikolay\">@nnikolay</a> Get feature correlation.</p>\n\n<p><a href=\"/adaubas\">@adaubas</a>, you say</p>\n\n<blockquote>\n  <p>hence tree methods should outperform logisitic regression in this new challenge</p>\n</blockquote>\n\n<p>If I'm understanding correctly, feature dependence/interaction causes colinearity in logistic regression, making weights unstable. Tree methods like random forest or gradient boosting is not affected by colinearity caused by feature dependence. Is this a correct elaboration of your statement?</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 713644,
      "author_name": "mpwolke",
      "author_url": "",
      "post_date": "01/08/2020 13:54:50",
      "content": "<p>J'ai oublié: Mes Félicitations M. Daubas  pour le votre accomplissement dans la dernière Competition (2eme place). Bonne chance.</p>",
      "votes": null,
      "replies": [
        {
          "id": 713914,
          "author_name": "adaubas",
          "author_url": "",
          "post_date": "01/08/2020 19:47:54",
          "content": "<p>merci et bonne chance aussi !</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "708518": "The subtitle of this Competition includes that expression : (and interactions!). Has anyone the idea of what kind of Interactions? How it changes the approaches?",
    "713288": "Hello,\nIn the first \"Categorical feature encoding challenge\", every feature were independant, there were no interaction between features. In this new challenge, features are not independant : hence tree methods should outperform logisitic regression in this new challenge (logistic regression won the first challenge).",
    "713612": "Thank you for the answer Aldparis . Now I understood the difference between the two Competitions.",
    "713644": "J'ai oublié: Mes Félicitations M. Daubas  pour le votre accomplissement dans la dernière Competition (2eme place). Bonne chance.",
    "713914": "merci et bonne chance aussi !",
    "717714": "adaubas how to check which features are dependent to the others?",
    "721791": "nnikolay Get feature correlation.\n\n@adaubas, you say\n\n&gt; hence tree methods should outperform logisitic regression in this new challenge\n\nIf I'm understanding correctly, feature dependence/interaction causes colinearity in logistic regression, making weights unstable. Tree methods like random forest or gradient boosting is not affected by colinearity caused by feature dependence. Is this a correct elaboration of your statement?"
  },
  "source": "meta"
}