{
  "id": 478408,
  "title": "Current KPIs used by Home Credit ",
  "url": "/competitions/home-credit-credit-risk-model-stability/discussion/478408",
  "author_name": "",
  "post_date": "2024-02-20T16:57:16.249057600Z",
  "votes": 1,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Hi,</p>\n<p>From the problem statement, I need to understand what are the current/required KPIs Home Credit firm is using to evaluate customers' likelihood of defaulting on loan repayment.</p>\n<p>Thanks!<br>\nKanishk</p>",
  "messages": [
    {
      "id": "2660542",
      "postDate": "02/20/2024 16:57:16",
      "content": "<p>Hi,</p>\n<p>From the problem statement, I need to understand what are the current/required KPIs Home Credit firm is using to evaluate customers' likelihood of defaulting on loan repayment.</p>\n<p>Thanks!<br>\nKanishk</p>",
      "rawMarkdown": "Hi,\n\nFrom the problem statement, I need to understand what are the current/required KPIs Home Credit firm is using to evaluate customers' likelihood of defaulting on loan repayment.\n\nThanks!\nKanishk",
      "votes": null
    },
    {
      "id": "2660564",
      "postDate": "02/20/2024 17:12:09",
      "content": "<p>Hi Kanishk,<br>\nnot sure I understand - do you mean KPIs for data scientists who develops models?</p>",
      "rawMarkdown": "Hi Kanishk,\nnot sure I understand - do you mean KPIs for data scientists who develops models?",
      "votes": null
    },
    {
      "id": "2660578",
      "postDate": "02/20/2024 17:22:31",
      "content": "<p>Hi Tomas, yes just need to understand what are the factors the firm is currently considering (if they're doing any data analysis on customer records) which can serve as input predictive factors for the models data scientists will develop.</p>\n<p>Hope this is a clear question :)</p>",
      "rawMarkdown": "Hi Tomas, yes just need to understand what are the factors the firm is currently considering (if they're doing any data analysis on customer records) which can serve as input predictive factors for the models data scientists will develop.\n\nHope this is a clear question :)",
      "votes": null
    },
    {
      "id": "2663082",
      "postDate": "02/22/2024 09:48:03",
      "content": "<p>No sure I have a clear answer, because it really depends on each project:</p>\n<ul>\n<li>build more predictive model (higher AUC) compare to previous generation</li>\n<li>build model that is good enough on each important customer segment (and here might be many different business priorities)</li>\n<li>restrictions on certain predictors (e.g., use \"expensive\" data only if they contribute significantly)</li>\n<li>stability of model and predictors</li>\n<li>requirements on interpretability (and alignment of \"business logic/common sense\" with model predictions), implementation complexity (this might limit what algorithm we can use)</li>\n<li>of course legal limitations (some data cannot be used)<br>\n…</li>\n</ul>",
      "rawMarkdown": "No sure I have a clear answer, because it really depends on each project:\n\n- build more predictive model (higher AUC) compare to previous generation\n- build model that is good enough on each important customer segment (and here might be many different business priorities)\n- restrictions on certain predictors (e.g., use \"expensive\" data only if they contribute significantly)\n- stability of model and predictors\n- requirements on interpretability (and alignment of \"business logic/common sense\" with model predictions), implementation complexity (this might limit what algorithm we can use)\n- of course legal limitations (some data cannot be used)\n...",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2660564,
      "author_name": "tomasjeline2",
      "author_url": "",
      "post_date": "02/20/2024 17:12:09",
      "content": "<p>Hi Kanishk,<br>\nnot sure I understand - do you mean KPIs for data scientists who develops models?</p>",
      "votes": null,
      "replies": [
        {
          "id": 2660578,
          "author_name": "kumanchi",
          "author_url": "",
          "post_date": "02/20/2024 17:22:31",
          "content": "<p>Hi Tomas, yes just need to understand what are the factors the firm is currently considering (if they're doing any data analysis on customer records) which can serve as input predictive factors for the models data scientists will develop.</p>\n<p>Hope this is a clear question :)</p>",
          "votes": null,
          "replies": [
            {
              "id": 2663082,
              "author_name": "tomasjeline2",
              "author_url": "",
              "post_date": "02/22/2024 09:48:03",
              "content": "<p>No sure I have a clear answer, because it really depends on each project:</p>\n<ul>\n<li>build more predictive model (higher AUC) compare to previous generation</li>\n<li>build model that is good enough on each important customer segment (and here might be many different business priorities)</li>\n<li>restrictions on certain predictors (e.g., use \"expensive\" data only if they contribute significantly)</li>\n<li>stability of model and predictors</li>\n<li>requirements on interpretability (and alignment of \"business logic/common sense\" with model predictions), implementation complexity (this might limit what algorithm we can use)</li>\n<li>of course legal limitations (some data cannot be used)<br>\n…</li>\n</ul>",
              "votes": null,
              "replies": []
            }
          ]
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2660542": "Hi,\n\nFrom the problem statement, I need to understand what are the current/required KPIs Home Credit firm is using to evaluate customers' likelihood of defaulting on loan repayment.\n\nThanks!\nKanishk",
    "2660564": "Hi Kanishk,\nnot sure I understand - do you mean KPIs for data scientists who develops models?",
    "2660578": "Hi Tomas, yes just need to understand what are the factors the firm is currently considering (if they're doing any data analysis on customer records) which can serve as input predictive factors for the models data scientists will develop.\n\nHope this is a clear question :)",
    "2663082": "No sure I have a clear answer, because it really depends on each project:\n\n- build more predictive model (higher AUC) compare to previous generation\n- build model that is good enough on each important customer segment (and here might be many different business priorities)\n- restrictions on certain predictors (e.g., use \"expensive\" data only if they contribute significantly)\n- stability of model and predictors\n- requirements on interpretability (and alignment of \"business logic/common sense\" with model predictions), implementation complexity (this might limit what algorithm we can use)\n- of course legal limitations (some data cannot be used)\n..."
  },
  "source": "meta"
}