{
  "id": 333000,
  "title": "Best Single Model",
  "url": "/competitions/amex-default-prediction/discussion/333000",
  "author_name": "zakopuro",
  "post_date": "2022-06-24T09:57:28.057000",
  "votes": 54,
  "comment_count": 32,
  "views": 0,
  "content": "<p>What is your current best single model?</p>\n<p>My</p>\n<pre><code>model : lightgbm\nfold : 5\nCV : 0.7953\nLB : 0.797\n</code></pre>\n<p>UPDATE 1</p>\n<pre><code>model : lightgbm\nfold : 5\nCV : 0.7976\nLB : 0.798\n</code></pre>",
  "messages": [
    {
      "id": 1831610,
      "postDate": "2022-06-24T09:57:28.057Z",
      "content": "<p>What is your current best single model?</p>\n<p>My</p>\n<pre><code>model : lightgbm\nfold : 5\nCV : 0.7953\nLB : 0.797\n</code></pre>\n<p>UPDATE 1</p>\n<pre><code>model : lightgbm\nfold : 5\nCV : 0.7976\nLB : 0.798\n</code></pre>",
      "rawMarkdown": "What is your current best single model?\n\nMy\n```\nmodel : lightgbm\nfold : 5\nCV : 0.7953\nLB : 0.797\n```\n\nUPDATE 1\n```\nmodel : lightgbm\nfold : 5\nCV : 0.7976\nLB : 0.798\n```",
      "votes": 52
    },
    {
      "id": 1834272,
      "postDate": "2022-06-26T19:13:31.407Z",
      "content": "<p>xgboost<br>\nCV: 0.7968<br>\nLB: 0.797</p>",
      "rawMarkdown": "xgboost\nCV: 0.7968\nLB: 0.797",
      "votes": 11,
      "replies": [
        {
          "id": 1850721,
          "postDate": "2022-07-10T16:46:53.290Z",
          "content": "<p>Excellent result! Which boosting type have you used: gbt or dart?</p>",
          "rawMarkdown": "Excellent result! Which boosting type have you used: gbt or dart?",
          "votes": 1
        }
      ]
    },
    {
      "id": 1835548,
      "postDate": "2022-06-27T20:29:18.937Z",
      "content": "<p>model : lightgbm<br>\nfold : 5<br>\nCV : 0.7969<br>\nLB : 0.798</p>",
      "rawMarkdown": "model : lightgbm\nfold : 5\nCV : 0.7969\nLB : 0.798",
      "votes": 7
    },
    {
      "id": 1834498,
      "postDate": "2022-06-27T01:44:22.790Z",
      "content": "<p>Just a theory, but perhaps the lack of sharing suggests that people are having a lot more luck with ensembles than single models ;)</p>",
      "rawMarkdown": "Just a theory, but perhaps the lack of sharing suggests that people are having a lot more luck with ensembles than single models ;)",
      "votes": 7,
      "replies": [
        {
          "id": 1834956,
          "postDate": "2022-06-27T10:56:12.987Z",
          "content": "<p>in most of the competitions, the ensemble of low correlated tuned single models are the winners.</p>",
          "rawMarkdown": "in most of the competitions, the ensemble of low correlated tuned single models are the winners.",
          "votes": 2
        },
        {
          "id": 1835252,
          "postDate": "2022-06-27T15:52:20.443Z",
          "content": "<p>There is a range though. Especially with larger training datasets (see e.g. <a href=\"https://www.kaggle.com/c/talkingdata-adtracking-fraud-detection\" target=\"_blank\">talkingdata</a> ensembling might not help much vs. great single models. Also, in some competitions single models are good enough for high silver or gold, while in others ensembles give a lower bound on competitive results.</p>",
          "rawMarkdown": "There is a range though. Especially with larger training datasets (see e.g. [talkingdata](https://www.kaggle.com/c/talkingdata-adtracking-fraud-detection) ensembling might not help much vs. great single models. Also, in some competitions single models are good enough for high silver or gold, while in others ensembles give a lower bound on competitive results.",
          "votes": 1
        },
        {
          "id": 1835844,
          "postDate": "2022-06-28T06:27:48.497Z",
          "content": "<p>Thanks for your explanation, Joe! </p>",
          "rawMarkdown": "Thanks for your explanation, Joe! ",
          "votes": 1
        }
      ]
    },
    {
      "id": 1834203,
      "postDate": "2022-06-26T18:08:58.603Z",
      "content": "<p>Thank you for sharing! And great job on the model result! <br>\nNot sure why people give upvotes but don't share their best model. Here's mine.</p>\n<pre><code>model: lightgbm\nfold: 5\nCV: 0.7966\nLB: 0.796\n</code></pre>",
      "rawMarkdown": "Thank you for sharing! And great job on the model result! \nNot sure why people give upvotes but don't share their best model. Here's mine.\n```\nmodel: lightgbm\nfold: 5\nCV: 0.7966\nLB: 0.796\n```",
      "votes": 5
    },
    {
      "id": 1834785,
      "postDate": "2022-06-27T07:27:59.403Z",
      "content": "<p>LightGBM<br>\nCV: 0.79732<br>\nLB: 0.797</p>",
      "rawMarkdown": "LightGBM\nCV: 0.79732\nLB: 0.797",
      "votes": 3
    },
    {
      "id": 1836815,
      "postDate": "2022-06-29T05:37:31.263Z",
      "content": "<p>fold : 1<br>\nCV：0.797<br>\nLB：0.798</p>",
      "rawMarkdown": "fold : 1\nCV：0.797\nLB：0.798\n",
      "votes": 4
    },
    {
      "id": 1834590,
      "postDate": "2022-06-27T04:25:49.053Z",
      "content": "<p><strong>UPDATES</strong></p>\n<p><strong>xgboost</strong><br>\nCV: 0.7969<br>\nLB: 0.798</p>\n<p><strong>LGBM</strong><br>\nCV: 0.7984<br>\nLB: 0.799</p>\n<p><strong>Cat</strong><br>\nCv=0.7963<br>\nLb=0.797</p>\n<p><strong>MLP</strong><br>\nCV: 0.79128<br>\nLB: 0.792</p>",
      "rawMarkdown": "**UPDATES**\n\n**xgboost**\nCV: 0.7969\nLB: 0.798\n\n**LGBM**\nCV: 0.7984\nLB: 0.799\n\n\n**Cat**\nCv=0.7963\nLb=0.797\n\n**MLP**\nCV: 0.79128\nLB: 0.792",
      "votes": 4
    },
    {
      "id": 1867299,
      "postDate": "2022-07-23T06:02:31.947Z",
      "content": "<p>model: lightgbm (seed 42)<br>\nnum_folds: 5<br>\ncv: 0.797883<br>\nlb: 0.799</p>\n<p>Did a decent amount of FE but it's proving incredibly hard for me to crack into 0.798 cv. Should I focus more on feature selection or other models like xgboost/catboost/NN's? If anyone has any suggestions please let me know. Thank you.</p>",
      "rawMarkdown": "model: lightgbm (seed 42)\nnum_folds: 5\ncv: 0.797883\nlb: 0.799\n\nDid a decent amount of FE but it's proving incredibly hard for me to crack into 0.798 cv. Should I focus more on feature selection or other models like xgboost/catboost/NN's? If anyone has any suggestions please let me know. Thank you.",
      "votes": 1,
      "replies": [
        {
          "id": 1867510,
          "postDate": "2022-07-23T09:30:32.053Z",
          "content": "<p>It's a dart or gbt?</p>",
          "rawMarkdown": "It's a dart or gbt?",
          "votes": 1
        },
        {
          "id": 1867545,
          "postDate": "2022-07-23T10:14:32.703Z",
          "content": "<p>Dart. I feel like I might have generated way too many correlated features or features with varying distributions in train/test. </p>",
          "rawMarkdown": "Dart. I feel like I might have generated way too many correlated features or features with varying distributions in train/test. "
        },
        {
          "id": 1867563,
          "postDate": "2022-07-23T10:41:57.723Z",
          "content": "<p>How many features do you have? Try another model and low colsample. </p>",
          "rawMarkdown": "How many features do you have? Try another model and low colsample. "
        },
        {
          "id": 1867582,
          "postDate": "2022-07-23T10:53:13.703Z",
          "content": "<p>Around 2200. Toying around with Olivier's null importance script right now. Will try lowering colsample as well, thank you for your input. Appreciate it.</p>",
          "rawMarkdown": "Around 2200. Toying around with Olivier's null importance script right now. Will try lowering colsample as well, thank you for your input. Appreciate it."
        },
        {
          "id": 1867715,
          "postDate": "2022-07-23T13:09:09.693Z",
          "content": "<p>I think it's time for me to switch to DART. I received CV on XGB gdt 0.7977, but on LB I got 0.797 ☹️ I really hoped to achieve 0.798 on LB, but time is running out</p>",
          "rawMarkdown": "I think it's time for me to switch to DART. I received CV on XGB gdt 0.7977, but on LB I got 0.797 ☹️ I really hoped to achieve 0.798 on LB, but time is running out"
        }
      ]
    },
    {
      "id": 1864579,
      "postDate": "2022-07-21T06:53:54.133Z",
      "content": "<p>xgboost(seed 42)<br>\n5folds<br>\nCV:0.797<br>\nLB:0.795</p>\n<p>Any explanation why its performing worse on public lb ?</p>",
      "rawMarkdown": "xgboost(seed 42)\n5folds\nCV:0.797\nLB:0.795\n\nAny explanation why its performing worse on public lb ?",
      "votes": 1,
      "replies": [
        {
          "id": 1864868,
          "postDate": "2022-07-21T10:54:21.480Z",
          "content": "<p>May be you accidentally preprocessed the training and test data in different ways. I had it so that I trained the model by preprocessing the training data, the CV was 0.7968, but then I noticed that the data preprocessing function was not suitable for the test data, changed it a little, ran predictions - the LB result was 0.793. Then I found out that I mixed the columns when changing the function for preprocessing in the test data. 🙈</p>",
          "rawMarkdown": "May be you accidentally preprocessed the training and test data in different ways. I had it so that I trained the model by preprocessing the training data, the CV was 0.7968, but then I noticed that the data preprocessing function was not suitable for the test data, changed it a little, ran predictions - the LB result was 0.793. Then I found out that I mixed the columns when changing the function for preprocessing in the test data. 🙈",
          "votes": 1
        },
        {
          "id": 1864929,
          "postDate": "2022-07-21T11:42:29.683Z",
          "content": "<p>Nice to see you here!<br>\nPlease check your cross-validation method.</p>",
          "rawMarkdown": "Nice to see you here!\nPlease check your cross-validation method.",
          "votes": 1
        },
        {
          "id": 1864967,
          "postDate": "2022-07-21T12:04:07.923Z",
          "content": "<p><a href=\"https://www.kaggle.com/xiaowangiiiii\" target=\"_blank\">@xiaowangiiiii</a> Yeah i am back after so long 😄</p>",
          "rawMarkdown": "@xiaowangiiiii Yeah i am back after so long 😄",
          "votes": 1
        },
        {
          "id": 1864968,
          "postDate": "2022-07-21T12:04:23.440Z",
          "rawMarkdown": "",
          "isDeleted": true
        }
      ]
    },
    {
      "id": 1850709,
      "postDate": "2022-07-10T16:33:57.723Z",
      "content": "<p><strong>update</strong></p>\n<p>XGB gbt<br>\n5 folds (seed 42)<br>\nCV: <strong>0.7977</strong><br>\nLB: 0.797</p>\n<p>LGB gbt<br>\n5 folds (seed 42)<br>\nCV: <strong>0.7977</strong><br>\nLB: 0.798</p>",
      "rawMarkdown": "**update**\n\nXGB gbt\n5 folds (seed 42)\nCV: **0.7977**\nLB: 0.797\n\nLGB gbt\n5 folds (seed 42)\nCV: **0.7977**\nLB: 0.798",
      "votes": 1
    },
    {
      "id": 1843625,
      "postDate": "2022-07-05T03:48:45.700Z",
      "content": "<p>model: lightgbm<br>\nfold: 5<br>\ncv: 0.7978<br>\nlb: 0.798</p>",
      "rawMarkdown": "model: lightgbm\nfold: 5\ncv: 0.7978\nlb: 0.798",
      "votes": 1
    },
    {
      "id": 1835737,
      "postDate": "2022-06-28T03:39:06.787Z",
      "content": "<p>model : lightgbm<br>\nfold : 5<br>\nCV : 0.7955<br>\nLB : 0.797</p>",
      "rawMarkdown": "model : lightgbm\nfold : 5\nCV : 0.7955\nLB : 0.797",
      "votes": 1
    },
    {
      "id": 1861638,
      "postDate": "2022-07-19T06:23:34.690Z",
      "content": "<p><strong>xgboost</strong><br>\n5folds<br>\n<strong>CV</strong>:0.7973<br>\n<strong>LB</strong>:0.798</p>",
      "rawMarkdown": "**xgboost**\n5folds\n**CV**:0.7973\n**LB**:0.798",
      "votes": 2,
      "replies": [
        {
          "id": 1862060,
          "postDate": "2022-07-19T12:35:01.903Z",
          "content": "<p>Cool! Which seed you used?</p>",
          "rawMarkdown": "Cool! Which seed you used?"
        },
        {
          "id": 1862079,
          "postDate": "2022-07-19T12:55:48.963Z",
          "content": "<p>Great work <a href=\"https://www.kaggle.com/tonymarkchris\" target=\"_blank\">@tonymarkchris</a> in xgb</p>",
          "rawMarkdown": "Great work @tonymarkchris in xgb"
        }
      ]
    },
    {
      "id": 1836873,
      "postDate": "2022-06-29T07:18:17.447Z",
      "content": "<p>model: xgboost <br>\nfold:5<br>\nCV:7963<br>\nLB:798</p>",
      "rawMarkdown": "model: xgboost \nfold:5\nCV:7963\nLB:798",
      "votes": 2,
      "replies": [
        {
          "id": 1837345,
          "postDate": "2022-06-29T13:59:52.060Z",
          "content": "<p>This is excellent👍 hope to achieve this </p>",
          "rawMarkdown": "This is excellent👍 hope to achieve this "
        },
        {
          "id": 1840057,
          "postDate": "2022-07-01T22:04:28.760Z",
          "content": "<p>great CV score</p>",
          "rawMarkdown": "great CV score"
        }
      ]
    },
    {
      "id": 1864905,
      "postDate": "2022-07-21T11:28:09.740Z",
      "content": "<p>omg nice   </p>",
      "rawMarkdown": "omg nice   "
    }
  ],
  "comments": [
    {
      "id": 1834272,
      "author_name": "raddar",
      "author_url": "",
      "post_date": "2022-06-26T19:13:31.407000",
      "content": "<p>xgboost<br>\nCV: 0.7968<br>\nLB: 0.797</p>",
      "votes": 11,
      "replies": [
        {
          "id": 1850721,
          "author_name": "Dmitry Uarov",
          "author_url": "",
          "post_date": "2022-07-10T16:46:53.290000",
          "content": "<p>Excellent result! Which boosting type have you used: gbt or dart?</p>",
          "votes": 1,
          "replies": []
        }
      ]
    },
    {
      "id": 1835548,
      "author_name": "Davut Polat",
      "author_url": "",
      "post_date": "2022-06-27T20:29:18.937000",
      "content": "<p>model : lightgbm<br>\nfold : 5<br>\nCV : 0.7969<br>\nLB : 0.798</p>",
      "votes": 7,
      "replies": []
    },
    {
      "id": 1834498,
      "author_name": "Joe Eddy",
      "author_url": "",
      "post_date": "2022-06-27T01:44:22.790000",
      "content": "<p>Just a theory, but perhaps the lack of sharing suggests that people are having a lot more luck with ensembles than single models ;)</p>",
      "votes": 7,
      "replies": [
        {
          "id": 1834956,
          "author_name": "1110Ra",
          "author_url": "",
          "post_date": "2022-06-27T10:56:12.987000",
          "content": "<p>in most of the competitions, the ensemble of low correlated tuned single models are the winners.</p>",
          "votes": 2,
          "replies": []
        },
        {
          "id": 1835252,
          "author_name": "Joe Eddy",
          "author_url": "",
          "post_date": "2022-06-27T15:52:20.443000",
          "content": "<p>There is a range though. Especially with larger training datasets (see e.g. <a href=\"https://www.kaggle.com/c/talkingdata-adtracking-fraud-detection\" target=\"_blank\">talkingdata</a> ensembling might not help much vs. great single models. Also, in some competitions single models are good enough for high silver or gold, while in others ensembles give a lower bound on competitive results.</p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 1835844,
          "author_name": "1110Ra",
          "author_url": "",
          "post_date": "2022-06-28T06:27:48.497000",
          "content": "<p>Thanks for your explanation, Joe! </p>",
          "votes": 1,
          "replies": []
        }
      ]
    },
    {
      "id": 1834203,
      "author_name": "Tonghui Li",
      "author_url": "",
      "post_date": "2022-06-26T18:08:58.603000",
      "content": "<p>Thank you for sharing! And great job on the model result! <br>\nNot sure why people give upvotes but don't share their best model. Here's mine.</p>\n<pre><code>model: lightgbm\nfold: 5\nCV: 0.7966\nLB: 0.796\n</code></pre>",
      "votes": 5,
      "replies": []
    },
    {
      "id": 1834785,
      "author_name": "1110Ra",
      "author_url": "",
      "post_date": "2022-06-27T07:27:59.403000",
      "content": "<p>LightGBM<br>\nCV: 0.79732<br>\nLB: 0.797</p>",
      "votes": 3,
      "replies": []
    },
    {
      "id": 1836815,
      "author_name": "Mingjie Wang",
      "author_url": "",
      "post_date": "2022-06-29T05:37:31.263000",
      "content": "<p>fold : 1<br>\nCV：0.797<br>\nLB：0.798</p>",
      "votes": 4,
      "replies": []
    },
    {
      "id": 1834590,
      "author_name": "Gaurav Rawat",
      "author_url": "",
      "post_date": "2022-06-27T04:25:49.053000",
      "content": "<p><strong>UPDATES</strong></p>\n<p><strong>xgboost</strong><br>\nCV: 0.7969<br>\nLB: 0.798</p>\n<p><strong>LGBM</strong><br>\nCV: 0.7984<br>\nLB: 0.799</p>\n<p><strong>Cat</strong><br>\nCv=0.7963<br>\nLb=0.797</p>\n<p><strong>MLP</strong><br>\nCV: 0.79128<br>\nLB: 0.792</p>",
      "votes": 4,
      "replies": []
    },
    {
      "id": 1867299,
      "author_name": "Arindam Baruah",
      "author_url": "",
      "post_date": "2022-07-23T06:02:31.947000",
      "content": "<p>model: lightgbm (seed 42)<br>\nnum_folds: 5<br>\ncv: 0.797883<br>\nlb: 0.799</p>\n<p>Did a decent amount of FE but it's proving incredibly hard for me to crack into 0.798 cv. Should I focus more on feature selection or other models like xgboost/catboost/NN's? If anyone has any suggestions please let me know. Thank you.</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1867510,
          "author_name": "Dmitry Uarov",
          "author_url": "",
          "post_date": "2022-07-23T09:30:32.053000",
          "content": "<p>It's a dart or gbt?</p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 1867545,
          "author_name": "Arindam Baruah",
          "author_url": "",
          "post_date": "2022-07-23T10:14:32.703000",
          "content": "<p>Dart. I feel like I might have generated way too many correlated features or features with varying distributions in train/test. </p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 1867563,
          "author_name": "Dmitry Uarov",
          "author_url": "",
          "post_date": "2022-07-23T10:41:57.723000",
          "content": "<p>How many features do you have? Try another model and low colsample. </p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 1867582,
          "author_name": "Arindam Baruah",
          "author_url": "",
          "post_date": "2022-07-23T10:53:13.703000",
          "content": "<p>Around 2200. Toying around with Olivier's null importance script right now. Will try lowering colsample as well, thank you for your input. Appreciate it.</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 1867715,
          "author_name": "Dmitry Uarov",
          "author_url": "",
          "post_date": "2022-07-23T13:09:09.693000",
          "content": "<p>I think it's time for me to switch to DART. I received CV on XGB gdt 0.7977, but on LB I got 0.797 ☹️ I really hoped to achieve 0.798 on LB, but time is running out</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1864579,
      "author_name": "kushal Agrawal",
      "author_url": "",
      "post_date": "2022-07-21T06:53:54.133000",
      "content": "<p>xgboost(seed 42)<br>\n5folds<br>\nCV:0.797<br>\nLB:0.795</p>\n<p>Any explanation why its performing worse on public lb ?</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1864868,
          "author_name": "Dmitry Uarov",
          "author_url": "",
          "post_date": "2022-07-21T10:54:21.480000",
          "content": "<p>May be you accidentally preprocessed the training and test data in different ways. I had it so that I trained the model by preprocessing the training data, the CV was 0.7968, but then I noticed that the data preprocessing function was not suitable for the test data, changed it a little, ran predictions - the LB result was 0.793. Then I found out that I mixed the columns when changing the function for preprocessing in the test data. 🙈</p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 1864929,
          "author_name": "Mingjie Wang",
          "author_url": "",
          "post_date": "2022-07-21T11:42:29.683000",
          "content": "<p>Nice to see you here!<br>\nPlease check your cross-validation method.</p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 1864967,
          "author_name": "kushal Agrawal",
          "author_url": "",
          "post_date": "2022-07-21T12:04:07.923000",
          "content": "<p><a href=\"https://www.kaggle.com/xiaowangiiiii\" target=\"_blank\">@xiaowangiiiii</a> Yeah i am back after so long 😄</p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 1864968,
          "author_name": "",
          "author_url": "",
          "post_date": "2022-07-21T12:04:23.440000",
          "content": "",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1850709,
      "author_name": "Dmitry Uarov",
      "author_url": "",
      "post_date": "2022-07-10T16:33:57.723000",
      "content": "<p><strong>update</strong></p>\n<p>XGB gbt<br>\n5 folds (seed 42)<br>\nCV: <strong>0.7977</strong><br>\nLB: 0.797</p>\n<p>LGB gbt<br>\n5 folds (seed 42)<br>\nCV: <strong>0.7977</strong><br>\nLB: 0.798</p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 1843625,
      "author_name": "ds.wook",
      "author_url": "",
      "post_date": "2022-07-05T03:48:45.700000",
      "content": "<p>model: lightgbm<br>\nfold: 5<br>\ncv: 0.7978<br>\nlb: 0.798</p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 1835737,
      "author_name": "Prateek Gupta",
      "author_url": "",
      "post_date": "2022-06-28T03:39:06.787000",
      "content": "<p>model : lightgbm<br>\nfold : 5<br>\nCV : 0.7955<br>\nLB : 0.797</p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 1861638,
      "author_name": "DJ_Xia",
      "author_url": "",
      "post_date": "2022-07-19T06:23:34.690000",
      "content": "<p><strong>xgboost</strong><br>\n5folds<br>\n<strong>CV</strong>:0.7973<br>\n<strong>LB</strong>:0.798</p>",
      "votes": 2,
      "replies": [
        {
          "id": 1862060,
          "author_name": "Dmitry Uarov",
          "author_url": "",
          "post_date": "2022-07-19T12:35:01.903000",
          "content": "<p>Cool! Which seed you used?</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 1862079,
          "author_name": "Gaurav Rawat",
          "author_url": "",
          "post_date": "2022-07-19T12:55:48.963000",
          "content": "<p>Great work <a href=\"https://www.kaggle.com/tonymarkchris\" target=\"_blank\">@tonymarkchris</a> in xgb</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1836873,
      "author_name": "mufeng",
      "author_url": "",
      "post_date": "2022-06-29T07:18:17.447000",
      "content": "<p>model: xgboost <br>\nfold:5<br>\nCV:7963<br>\nLB:798</p>",
      "votes": 2,
      "replies": [
        {
          "id": 1837345,
          "author_name": "Gaurav Rawat",
          "author_url": "",
          "post_date": "2022-06-29T13:59:52.060000",
          "content": "<p>This is excellent👍 hope to achieve this </p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 1840057,
          "author_name": "stallone",
          "author_url": "",
          "post_date": "2022-07-01T22:04:28.760000",
          "content": "<p>great CV score</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1864905,
      "author_name": "Rahul Prasad M.",
      "author_url": "",
      "post_date": "2022-07-21T11:28:09.740000",
      "content": "<p>omg nice   </p>",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1831610": "What is your current best single model?\n\nMy\n```\nmodel : lightgbm\nfold : 5\nCV : 0.7953\nLB : 0.797\n```\n\nUPDATE 1\n```\nmodel : lightgbm\nfold : 5\nCV : 0.7976\nLB : 0.798\n```",
    "1834272": "xgboost\nCV: 0.7968\nLB: 0.797",
    "1835548": "model : lightgbm\nfold : 5\nCV : 0.7969\nLB : 0.798",
    "1834498": "Just a theory, but perhaps the lack of sharing suggests that people are having a lot more luck with ensembles than single models ;)",
    "1834203": "Thank you for sharing! And great job on the model result! \nNot sure why people give upvotes but don't share their best model. Here's mine.\n```\nmodel: lightgbm\nfold: 5\nCV: 0.7966\nLB: 0.796\n```",
    "1834785": "LightGBM\nCV: 0.79732\nLB: 0.797",
    "1836815": "fold : 1\nCV：0.797\nLB：0.798\n",
    "1834590": "**UPDATES**\n\n**xgboost**\nCV: 0.7969\nLB: 0.798\n\n**LGBM**\nCV: 0.7984\nLB: 0.799\n\n\n**Cat**\nCv=0.7963\nLb=0.797\n\n**MLP**\nCV: 0.79128\nLB: 0.792",
    "1867299": "model: lightgbm (seed 42)\nnum_folds: 5\ncv: 0.797883\nlb: 0.799\n\nDid a decent amount of FE but it's proving incredibly hard for me to crack into 0.798 cv. Should I focus more on feature selection or other models like xgboost/catboost/NN's? If anyone has any suggestions please let me know. Thank you.",
    "1864579": "xgboost(seed 42)\n5folds\nCV:0.797\nLB:0.795\n\nAny explanation why its performing worse on public lb ?",
    "1850709": "**update**\n\nXGB gbt\n5 folds (seed 42)\nCV: **0.7977**\nLB: 0.797\n\nLGB gbt\n5 folds (seed 42)\nCV: **0.7977**\nLB: 0.798",
    "1843625": "model: lightgbm\nfold: 5\ncv: 0.7978\nlb: 0.798",
    "1835737": "model : lightgbm\nfold : 5\nCV : 0.7955\nLB : 0.797",
    "1861638": "**xgboost**\n5folds\n**CV**:0.7973\n**LB**:0.798",
    "1836873": "model: xgboost \nfold:5\nCV:7963\nLB:798",
    "1864905": "omg nice   "
  }
}