{
  "id": 339389,
  "title": "AMEX REFERENCES",
  "url": "/competitions/amex-default-prediction/discussion/339389",
  "author_name": "",
  "post_date": "2022-07-24T15:01:32.454543400Z",
  "votes": 19,
  "comment_count": 4,
  "views": 0,
  "content": "<p>Hello,<br>\nI just want to share references, I found about amex's credit default model. I was wondering if someone would share there thoughts about this?</p>\n<p>Sequential Deep Learning for Credit Risk Monitoring with<br>\nTabular Financial Data<br>\n<a href=\"https://arxiv.org/pdf/2012.15330.pdf\" target=\"_blank\">https://arxiv.org/pdf/2012.15330.pdf</a></p>\n<p>Dr. Di Xu, American Express, Machine Learning at American Express<br>\n<a href=\"https://www.youtube.com/watch?v=fiSfB74yvlk\" target=\"_blank\">https://www.youtube.com/watch?v=fiSfB74yvlk</a><br>\n(AMEX MODEL is somewhere after the 20 minute mark)</p>",
  "messages": [
    {
      "id": "1869183",
      "postDate": "07/24/2022 15:01:32",
      "content": "<p>Hello,<br>\nI just want to share references, I found about amex's credit default model. I was wondering if someone would share there thoughts about this?</p>\n<p>Sequential Deep Learning for Credit Risk Monitoring with<br>\nTabular Financial Data<br>\n<a href=\"https://arxiv.org/pdf/2012.15330.pdf\" target=\"_blank\">https://arxiv.org/pdf/2012.15330.pdf</a></p>\n<p>Dr. Di Xu, American Express, Machine Learning at American Express<br>\n<a href=\"https://www.youtube.com/watch?v=fiSfB74yvlk\" target=\"_blank\">https://www.youtube.com/watch?v=fiSfB74yvlk</a><br>\n(AMEX MODEL is somewhere after the 20 minute mark)</p>",
      "rawMarkdown": "Hello,\nI just want to share references, I found about amex's credit default model. I was wondering if someone would share there thoughts about this?\n\n\nSequential Deep Learning for Credit Risk Monitoring with\nTabular Financial Data\nhttps://arxiv.org/pdf/2012.15330.pdf\n\nDr. Di Xu, American Express, Machine Learning at American Express\nhttps://www.youtube.com/watch?v=fiSfB74yvlk\n(AMEX MODEL is somewhere after the 20 minute mark)",
      "votes": null
    },
    {
      "id": "1871485",
      "postDate": "07/26/2022 09:48:20",
      "content": "<p>I think that the part about clipping the features before using a GBM model can be really interesting considering that this comes from the actual domain of the competition.<br>\nI will probably give it a try. <br>\nIn general it is great to see how a large financial institution like American Express is using \"kagglers\" machine learning models to improve their credit risk monitoring.</p>\n<p>Thanks again for sharing!</p>\n<p>The Devastator.</p>",
      "rawMarkdown": "I think that the part about clipping the features before using a GBM model can be really interesting considering that this comes from the actual domain of the competition.\nI will probably give it a try. \nIn general it is great to see how a large financial institution like American Express is using \"kagglers\" machine learning models to improve their credit risk monitoring.\n\nThanks again for sharing!\n\nThe Devastator.",
      "votes": null
    },
    {
      "id": "1871905",
      "postDate": "07/26/2022 14:49:58",
      "content": "<p>clipping features does sound unique, but applying it to time series features is weird in a sense since the mean of some features increase/decrease over time, don't you think?</p>",
      "rawMarkdown": "clipping features does sound unique, but applying it to time series features is weird in a sense since the mean of some features increase/decrease over time, don't you think?",
      "votes": null
    },
    {
      "id": "1872353",
      "postDate": "07/26/2022 21:08:56",
      "content": "<p>Thanks for sharing, does those papers explain the categories below or some examples? </p>\n<pre><code>D_* = Delinquency variables\nS_* = Spend variables\nP_* = Payment variables\nB_* = Balance variables\nR_* = Risk variables\n</code></pre>",
      "rawMarkdown": "Thanks for sharing, does those papers explain the categories below or some examples? \n\n    D_* = Delinquency variables\n    S_* = Spend variables\n    P_* = Payment variables\n    B_* = Balance variables\n    R_* = Risk variables",
      "votes": null
    },
    {
      "id": "1872493",
      "postDate": "07/27/2022 03:02:05",
      "content": "<p>The paper outlines the deep learning approach their current/past model is using, the features are mentioned briefly because of privacy concerns</p>",
      "rawMarkdown": "The paper outlines the deep learning approach their current/past model is using, the features are mentioned briefly because of privacy concerns",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1871485,
      "author_name": "thedevastator",
      "author_url": "",
      "post_date": "07/26/2022 09:48:20",
      "content": "<p>I think that the part about clipping the features before using a GBM model can be really interesting considering that this comes from the actual domain of the competition.<br>\nI will probably give it a try. <br>\nIn general it is great to see how a large financial institution like American Express is using \"kagglers\" machine learning models to improve their credit risk monitoring.</p>\n<p>Thanks again for sharing!</p>\n<p>The Devastator.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1871905,
          "author_name": "tarrasque9",
          "author_url": "",
          "post_date": "07/26/2022 14:49:58",
          "content": "<p>clipping features does sound unique, but applying it to time series features is weird in a sense since the mean of some features increase/decrease over time, don't you think?</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1872353,
      "author_name": "marcusrb",
      "author_url": "",
      "post_date": "07/26/2022 21:08:56",
      "content": "<p>Thanks for sharing, does those papers explain the categories below or some examples? </p>\n<pre><code>D_* = Delinquency variables\nS_* = Spend variables\nP_* = Payment variables\nB_* = Balance variables\nR_* = Risk variables\n</code></pre>",
      "votes": null,
      "replies": [
        {
          "id": 1872493,
          "author_name": "tarrasque9",
          "author_url": "",
          "post_date": "07/27/2022 03:02:05",
          "content": "<p>The paper outlines the deep learning approach their current/past model is using, the features are mentioned briefly because of privacy concerns</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1869183": "Hello,\nI just want to share references, I found about amex's credit default model. I was wondering if someone would share there thoughts about this?\n\n\nSequential Deep Learning for Credit Risk Monitoring with\nTabular Financial Data\nhttps://arxiv.org/pdf/2012.15330.pdf\n\nDr. Di Xu, American Express, Machine Learning at American Express\nhttps://www.youtube.com/watch?v=fiSfB74yvlk\n(AMEX MODEL is somewhere after the 20 minute mark)",
    "1871485": "I think that the part about clipping the features before using a GBM model can be really interesting considering that this comes from the actual domain of the competition.\nI will probably give it a try. \nIn general it is great to see how a large financial institution like American Express is using \"kagglers\" machine learning models to improve their credit risk monitoring.\n\nThanks again for sharing!\n\nThe Devastator.",
    "1871905": "clipping features does sound unique, but applying it to time series features is weird in a sense since the mean of some features increase/decrease over time, don't you think?",
    "1872353": "Thanks for sharing, does those papers explain the categories below or some examples? \n\n    D_* = Delinquency variables\n    S_* = Spend variables\n    P_* = Payment variables\n    B_* = Balance variables\n    R_* = Risk variables",
    "1872493": "The paper outlines the deep learning approach their current/past model is using, the features are mentioned briefly because of privacy concerns"
  },
  "source": "meta"
}