{
  "id": 339191,
  "title": "Performance of models build with categorical columns only",
  "url": "/competitions/amex-default-prediction/discussion/339191",
  "author_name": "Ravi Singh",
  "post_date": "2022-07-23T17:57:34.141000",
  "votes": 4,
  "comment_count": 5,
  "views": 0,
  "content": "<p>I applied XGboost and neural network on the categorical columns only. The score was 0.172.</p>\n<p>Although, the test set accuracy in my data set was around 74 percent</p>",
  "messages": [
    {
      "id": 1868110,
      "postDate": "2022-07-23T17:57:34.143Z",
      "content": "<p>I applied XGboost and neural network on the categorical columns only. The score was 0.172.</p>\n<p>Although, the test set accuracy in my data set was around 74 percent</p>",
      "rawMarkdown": "I applied XGboost and neural network on the categorical columns only. The score was 0.172.\n\nAlthough, the test set accuracy in my data set was around 74 percent",
      "votes": 4
    },
    {
      "id": 1871423,
      "postDate": "2022-07-26T09:03:26.587Z",
      "content": "<p>This might be an interesting experiment. <br>\nTry to encode the categorical variables with frequency encoding.</p>\n<p>The Devastator.</p>",
      "rawMarkdown": "This might be an interesting experiment. \nTry to encode the categorical variables with frequency encoding.\n\nThe Devastator.\n",
      "votes": 1,
      "replies": [
        {
          "id": 1871631,
          "postDate": "2022-07-26T11:41:50.223Z",
          "content": "<p>Thanks for suggesting Devastator. There are some nominal features as well; will it cause an issue? I will try to group them based on the frequency and see/share what happens.</p>\n<p>I was also thinking of transposing the months and treating them as individual features.</p>",
          "rawMarkdown": "Thanks for suggesting Devastator. There are some nominal features as well; will it cause an issue? I will try to group them based on the frequency and see/share what happens.\n\nI was also thinking of transposing the months and treating them as individual features."
        }
      ]
    },
    {
      "id": 1871626,
      "postDate": "2022-07-26T11:39:00.580Z",
      "content": "<p>With 13 categorical features. Not including the date: .486</p>",
      "rawMarkdown": "With 13 categorical features. Not including the date: .486",
      "votes": 2
    },
    {
      "id": 1871629,
      "postDate": "2022-07-26T11:41:19.587Z",
      "content": "<p>Yes Tord, I have not used date and D_63 and D_64. I will be one hot encode them and share the results in a while.</p>",
      "rawMarkdown": "Yes Tord, I have not used date and D_63 and D_64. I will be one hot encode them and share the results in a while."
    },
    {
      "id": 1871628,
      "postDate": "2022-07-26T11:39:55.297Z",
      "rawMarkdown": "",
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 1871423,
      "author_name": "The Devastator",
      "author_url": "",
      "post_date": "2022-07-26T09:03:26.587000",
      "content": "<p>This might be an interesting experiment. <br>\nTry to encode the categorical variables with frequency encoding.</p>\n<p>The Devastator.</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1871631,
          "author_name": "Ravi Singh",
          "author_url": "",
          "post_date": "2022-07-26T11:41:50.223000",
          "content": "<p>Thanks for suggesting Devastator. There are some nominal features as well; will it cause an issue? I will try to group them based on the frequency and see/share what happens.</p>\n<p>I was also thinking of transposing the months and treating them as individual features.</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1871626,
      "author_name": "Tord Malmgren",
      "author_url": "",
      "post_date": "2022-07-26T11:39:00.580000",
      "content": "<p>With 13 categorical features. Not including the date: .486</p>",
      "votes": 2,
      "replies": []
    },
    {
      "id": 1871629,
      "author_name": "Ravi Singh",
      "author_url": "",
      "post_date": "2022-07-26T11:41:19.587000",
      "content": "<p>Yes Tord, I have not used date and D_63 and D_64. I will be one hot encode them and share the results in a while.</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1871628,
      "author_name": "",
      "author_url": "",
      "post_date": "2022-07-26T11:39:55.297000",
      "content": "",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1868110": "I applied XGboost and neural network on the categorical columns only. The score was 0.172.\n\nAlthough, the test set accuracy in my data set was around 74 percent",
    "1871423": "This might be an interesting experiment. \nTry to encode the categorical variables with frequency encoding.\n\nThe Devastator.\n",
    "1871626": "With 13 categorical features. Not including the date: .486",
    "1871629": "Yes Tord, I have not used date and D_63 and D_64. I will be one hot encode them and share the results in a while.",
    "1871628": ""
  }
}