{
  "id": 328846,
  "title": "🥇 Model Performance Ranking",
  "url": "/competitions/amex-default-prediction/discussion/328846",
  "author_name": "",
  "post_date": "2022-06-03T08:40:57.202570800Z",
  "votes": 35,
  "comment_count": 8,
  "views": 0,
  "content": "<p>Ensembling different models is one of key factors to win this competition.<br>\nThis notebook is Top model ranking based on public score.</p>\n<p>It could be a guideline to find area to improve score.<br>\nI try to update this list regularly. Hope it is helpful.</p>\n<p>So far, boosting based models beat neural network.<br>\nAnd, most of the notebook is kindly shared by GM.</p>\n<p>Caution: Each model may have difference dataset, features and optimization level. <br>\nLower score model may not mean that model is not performing. It may show more room for improvment.</p>\n<table>\n<thead>\n<tr>\n<th>Rank</th>\n<th>Model</th>\n<th>Public Score</th>\n<th>Notebooks</th>\n</tr>\n</thead>\n<tbody>\n<tr>\n<td>1</td>\n<td>XGBoost</td>\n<td>0.795  🥇</td>\n<td><a href=\"https://www.kaggle.com/code/jiweiliu/rapids-cudf-feature-engineering-xgb\" target=\"_blank\">RAPIDS cudf Feature Engineering + XGB</a> (Jiwei Liu)   <br> <a href=\"https://www.kaggle.com/code/cdeotte/xgboost-starter-0-793\" target=\"_blank\">xgboost-starter-0-793</a> (Chris Deotte)</td>\n</tr>\n<tr>\n<td>2</td>\n<td>Catboost</td>\n<td>0.794  🥇</td>\n<td><a href=\"https://www.kaggle.com/code/jiweiliu/amex-catboost-rounding-trick\" target=\"_blank\">amex-catboost-rounding-trick</a> (Jiwei Liu)   <br><a href=\"https://www.kaggle.com/code/huseyincot/amex-catboost-0-793\" target=\"_blank\">amex-catboost-0-793</a> (huseyincotel)</td>\n</tr>\n<tr>\n<td>3▲</td>\n<td>LightGBM</td>\n<td>0.793 🥇</td>\n<td><a href=\"https://www.kaggle.com/code/lucasmorin/amex-lgbm-features-eng\" target=\"_blank\">amex-lgbm-features-eng</a> (Lucas Morin) <br> <a href=\"https://www.kaggle.com/code/ambrosm/amex-lightgbm-quickstart\" target=\"_blank\">amex-lightgbm-quickstart</a> (AmbrosM) <br> <a href=\"https://www.kaggle.com/code/kellibelcher/amex-default-prediction-eda-lgbm-baseline\" target=\"_blank\">AMEX Default Prediction EDA &amp; LGBM Baseline</a> (Kelli Belcher)</td>\n</tr>\n<tr>\n<td>4▼</td>\n<td>Py-Boost</td>\n<td>0.792</td>\n<td><a href=\"https://www.kaggle.com/code/btbpanda/fast-metric-and-py-boost-baseline\" target=\"_blank\">fast-metric-and-py-boost-baseline</a> (Btbpanda)</td>\n</tr>\n<tr>\n<td>5</td>\n<td>Keras Dense</td>\n<td>0.790</td>\n<td><a href=\"https://www.kaggle.com/code/ambrosm/amex-keras-quickstart-1-training\" target=\"_blank\">amex-keras-quickstart-1-training</a> (AmbrosM)</td>\n</tr>\n<tr>\n<td>6</td>\n<td>TensorFlow Transformer</td>\n<td>0.789</td>\n<td><a href=\"https://www.kaggle.com/code/cdeotte/tensorflow-transformer-0-790\" target=\"_blank\">tensorflow-transformer-0-790</a> (Chris Deotte)</td>\n</tr>\n<tr>\n<td>7</td>\n<td>TensorFlow GRU</td>\n<td>0.789</td>\n<td><a href=\"https://www.kaggle.com/code/cdeotte/tensorflow-gru-starter-0-790\" target=\"_blank\">tensorflow-gru-starter-0-790</a> (Chris Deotte)</td>\n</tr>\n<tr>\n<td>8</td>\n<td>PyTorch NN</td>\n<td>0.787</td>\n<td><a href=\"https://www.kaggle.com/code/voix97/amex-pytorch-nn-training\" target=\"_blank\">AMEX PyTorch NN: Training</a> (Xiang Sheng)</td>\n</tr>\n<tr>\n<td>9▲</td>\n<td>PyTorch GRU</td>\n<td>0.786</td>\n<td><a href=\"https://www.kaggle.com/code/voix97/amex-pytorch-gru-training\" target=\"_blank\">AMEX PyTorch GRU: Training</a> (Xiang Sheng)</td>\n</tr>\n<tr>\n<td>10▼</td>\n<td>WoE</td>\n<td>0.757</td>\n<td><a href=\"https://www.kaggle.com/code/gopidurgaprasad/amex-credit-score-model\" target=\"_blank\">amex-credit-score-model</a>  (Gopi Durgaprasad)</td>\n</tr>\n<tr>\n<td>11▼</td>\n<td>Random Forest</td>\n<td>0.715</td>\n<td><a href=\"https://www.kaggle.com/code/bgmello/best-correlated-features-with-low-correlation\" target=\"_blank\">best-correlated-features-with-low-correlation</a> (Bruno Gorresen Mello)</td>\n</tr>\n</tbody>\n</table>\n<p>※ As long as it is not combined with another model, it will be in this ranking. <br>\n※ Sorry if I miss any models. Please let me know, I will correct it.<br>\n※ Please share your private share, I plan to add into this ranking.</p>\n<p>※ Updated: 2022-06-08</p>",
  "messages": [
    {
      "id": "1810078",
      "postDate": "06/03/2022 08:40:57",
      "content": "<p>Ensembling different models is one of key factors to win this competition.<br>\nThis notebook is Top model ranking based on public score.</p>\n<p>It could be a guideline to find area to improve score.<br>\nI try to update this list regularly. Hope it is helpful.</p>\n<p>So far, boosting based models beat neural network.<br>\nAnd, most of the notebook is kindly shared by GM.</p>\n<p>Caution: Each model may have difference dataset, features and optimization level. <br>\nLower score model may not mean that model is not performing. It may show more room for improvment.</p>\n<table>\n<thead>\n<tr>\n<th>Rank</th>\n<th>Model</th>\n<th>Public Score</th>\n<th>Notebooks</th>\n</tr>\n</thead>\n<tbody>\n<tr>\n<td>1</td>\n<td>XGBoost</td>\n<td>0.795  🥇</td>\n<td><a href=\"https://www.kaggle.com/code/jiweiliu/rapids-cudf-feature-engineering-xgb\" target=\"_blank\">RAPIDS cudf Feature Engineering + XGB</a> (Jiwei Liu)   <br> <a href=\"https://www.kaggle.com/code/cdeotte/xgboost-starter-0-793\" target=\"_blank\">xgboost-starter-0-793</a> (Chris Deotte)</td>\n</tr>\n<tr>\n<td>2</td>\n<td>Catboost</td>\n<td>0.794  🥇</td>\n<td><a href=\"https://www.kaggle.com/code/jiweiliu/amex-catboost-rounding-trick\" target=\"_blank\">amex-catboost-rounding-trick</a> (Jiwei Liu)   <br><a href=\"https://www.kaggle.com/code/huseyincot/amex-catboost-0-793\" target=\"_blank\">amex-catboost-0-793</a> (huseyincotel)</td>\n</tr>\n<tr>\n<td>3▲</td>\n<td>LightGBM</td>\n<td>0.793 🥇</td>\n<td><a href=\"https://www.kaggle.com/code/lucasmorin/amex-lgbm-features-eng\" target=\"_blank\">amex-lgbm-features-eng</a> (Lucas Morin) <br> <a href=\"https://www.kaggle.com/code/ambrosm/amex-lightgbm-quickstart\" target=\"_blank\">amex-lightgbm-quickstart</a> (AmbrosM) <br> <a href=\"https://www.kaggle.com/code/kellibelcher/amex-default-prediction-eda-lgbm-baseline\" target=\"_blank\">AMEX Default Prediction EDA &amp; LGBM Baseline</a> (Kelli Belcher)</td>\n</tr>\n<tr>\n<td>4▼</td>\n<td>Py-Boost</td>\n<td>0.792</td>\n<td><a href=\"https://www.kaggle.com/code/btbpanda/fast-metric-and-py-boost-baseline\" target=\"_blank\">fast-metric-and-py-boost-baseline</a> (Btbpanda)</td>\n</tr>\n<tr>\n<td>5</td>\n<td>Keras Dense</td>\n<td>0.790</td>\n<td><a href=\"https://www.kaggle.com/code/ambrosm/amex-keras-quickstart-1-training\" target=\"_blank\">amex-keras-quickstart-1-training</a> (AmbrosM)</td>\n</tr>\n<tr>\n<td>6</td>\n<td>TensorFlow Transformer</td>\n<td>0.789</td>\n<td><a href=\"https://www.kaggle.com/code/cdeotte/tensorflow-transformer-0-790\" target=\"_blank\">tensorflow-transformer-0-790</a> (Chris Deotte)</td>\n</tr>\n<tr>\n<td>7</td>\n<td>TensorFlow GRU</td>\n<td>0.789</td>\n<td><a href=\"https://www.kaggle.com/code/cdeotte/tensorflow-gru-starter-0-790\" target=\"_blank\">tensorflow-gru-starter-0-790</a> (Chris Deotte)</td>\n</tr>\n<tr>\n<td>8</td>\n<td>PyTorch NN</td>\n<td>0.787</td>\n<td><a href=\"https://www.kaggle.com/code/voix97/amex-pytorch-nn-training\" target=\"_blank\">AMEX PyTorch NN: Training</a> (Xiang Sheng)</td>\n</tr>\n<tr>\n<td>9▲</td>\n<td>PyTorch GRU</td>\n<td>0.786</td>\n<td><a href=\"https://www.kaggle.com/code/voix97/amex-pytorch-gru-training\" target=\"_blank\">AMEX PyTorch GRU: Training</a> (Xiang Sheng)</td>\n</tr>\n<tr>\n<td>10▼</td>\n<td>WoE</td>\n<td>0.757</td>\n<td><a href=\"https://www.kaggle.com/code/gopidurgaprasad/amex-credit-score-model\" target=\"_blank\">amex-credit-score-model</a>  (Gopi Durgaprasad)</td>\n</tr>\n<tr>\n<td>11▼</td>\n<td>Random Forest</td>\n<td>0.715</td>\n<td><a href=\"https://www.kaggle.com/code/bgmello/best-correlated-features-with-low-correlation\" target=\"_blank\">best-correlated-features-with-low-correlation</a> (Bruno Gorresen Mello)</td>\n</tr>\n</tbody>\n</table>\n<p>※ As long as it is not combined with another model, it will be in this ranking. <br>\n※ Sorry if I miss any models. Please let me know, I will correct it.<br>\n※ Please share your private share, I plan to add into this ranking.</p>\n<p>※ Updated: 2022-06-08</p>",
      "rawMarkdown": "Ensembling different models is one of key factors to win this competition.\nThis notebook is Top model ranking based on public score.\n\nIt could be a guideline to find area to improve score.\nI try to update this list regularly. Hope it is helpful.\n\nSo far, boosting based models beat neural network.\nAnd, most of the notebook is kindly shared by GM.\n\nCaution: Each model may have difference dataset, features and optimization level. \nLower score model may not mean that model is not performing. It may show more room for improvment.\n\n| Rank | Model | Public Score | Notebooks | \n| :---: | :---: | :---: | \n| 1 | XGBoost | 0.795  🥇 |  [RAPIDS cudf Feature Engineering + XGB][xgb2] (Jiwei Liu)   <br> [xgboost-starter-0-793][xgb1] (Chris Deotte)   | \n| 2 | Catboost | 0.794  🥇 |  [amex-catboost-rounding-trick][2.1] (Jiwei Liu)   <br>[amex-catboost-0-793][2.2] (huseyincotel)  | \n| 3▲| LightGBM | 0.793 🥇|   [amex-lgbm-features-eng][4.2] (Lucas Morin) <br> [amex-lightgbm-quickstart][4.1] (AmbrosM) <br> [AMEX Default Prediction EDA & LGBM Baseline][4.3] (Kelli Belcher)  | \n| 4▼| Py-Boost | 0.792 |  [fast-metric-and-py-boost-baseline][3] (Btbpanda)    | \n| 5 | Keras Dense | 0.790 |   [amex-keras-quickstart-1-training][5] (AmbrosM)   | \n| 6 | TensorFlow Transformer | 0.789 |   [tensorflow-transformer-0-790][6] (Chris Deotte)  | \n| 7 | TensorFlow GRU | 0.789 |  [tensorflow-gru-starter-0-790][7] (Chris Deotte)    | \n| 8 | PyTorch NN | 0.787 | [AMEX PyTorch NN: Training][PTN] (Xiang Sheng)    | \n| 9▲ | PyTorch GRU |  0.786 | [AMEX PyTorch GRU: Training][PTG] (Xiang Sheng)    | \n| 10▼ | WoE | 0.757 |  [amex-credit-score-model][9]  (Gopi Durgaprasad)    | \n| 11▼ | Random Forest | 0.715 |   [best-correlated-features-with-low-correlation][10] (Bruno Gorresen Mello)  | \n\n※ As long as it is not combined with another model, it will be in this ranking. \n※ Sorry if I miss any models. Please let me know, I will correct it.\n※ Please share your private share, I plan to add into this ranking.\n\n※ Updated: 2022-06-08\n\n[xgb1]: https://www.kaggle.com/code/cdeotte/xgboost-starter-0-793\n[xgb2]: https://www.kaggle.com/code/jiweiliu/rapids-cudf-feature-engineering-xgb\n[2.1]: https://www.kaggle.com/code/jiweiliu/amex-catboost-rounding-trick\n[2.2]: https://www.kaggle.com/code/huseyincot/amex-catboost-0-793\n[3]: https://www.kaggle.com/code/btbpanda/fast-metric-and-py-boost-baseline\n[4.1]: https://www.kaggle.com/code/ambrosm/amex-lightgbm-quickstart\n[4.2]: https://www.kaggle.com/code/lucasmorin/amex-lgbm-features-eng\n[4.3]: https://www.kaggle.com/code/kellibelcher/amex-default-prediction-eda-lgbm-baseline\n[5]: https://www.kaggle.com/code/ambrosm/amex-keras-quickstart-1-training\n[6]: https://www.kaggle.com/code/cdeotte/tensorflow-transformer-0-790\n[7]: https://www.kaggle.com/code/cdeotte/tensorflow-gru-starter-0-790\n[PTN]: https://www.kaggle.com/code/voix97/amex-pytorch-nn-training\n[PTG]: https://www.kaggle.com/code/voix97/amex-pytorch-gru-training\n[9]: https://www.kaggle.com/code/gopidurgaprasad/amex-credit-score-model\n[10]: https://www.kaggle.com/code/bgmello/best-correlated-features-with-low-correlation",
      "votes": null
    },
    {
      "id": "1810098",
      "postDate": "06/03/2022 09:14:34",
      "content": "<p>lightautoml one is actually an ensemble of a Logistic Regression, a LightGBM, and a Catboost.</p>\n<p>From the log:</p>\n<pre><code>Final prediction for new objects (level 0) = \n     0.11480 * (5 averaged models Lvl_0_Pipe_0_Mod_0_LinearL2) +\n     0.33581 * (5 averaged models Lvl_0_Pipe_1_Mod_0_LightGBM) +\n     0.54940 * (5 averaged models Lvl_0_Pipe_1_Mod_1_CatBoost) \n</code></pre>",
      "rawMarkdown": "lightautoml one is actually an ensemble of a Logistic Regression, a LightGBM, and a Catboost.\n\nFrom the log:\n```\nFinal prediction for new objects (level 0) = \n\t 0.11480 * (5 averaged models Lvl_0_Pipe_0_Mod_0_LinearL2) +\n\t 0.33581 * (5 averaged models Lvl_0_Pipe_1_Mod_0_LightGBM) +\n\t 0.54940 * (5 averaged models Lvl_0_Pipe_1_Mod_1_CatBoost) \n```",
      "votes": null
    },
    {
      "id": "1810274",
      "postDate": "06/03/2022 12:07:20",
      "content": "<p>Thank you for your comment.<br>\nI agree and removed automl from the list.</p>",
      "rawMarkdown": "Thank you for your comment.\nI agree and removed automl from the list.",
      "votes": null
    },
    {
      "id": "1810409",
      "postDate": "06/03/2022 14:34:49",
      "content": "<p>Nice list. You're missing LGBM at 792 from AmbrosM <a href=\"https://www.kaggle.com/code/ambrosm/amex-lightgbm-quickstart\" target=\"_blank\">here</a>. (And Lucas Morin created an LGBM 792 also <a href=\"https://www.kaggle.com/code/lucasmorin/amex-lgbm-features-eng\" target=\"_blank\">here</a> but we don't need two LGBM in the list)</p>",
      "rawMarkdown": "Nice list. You're missing LGBM at 792 from AmbrosM [here][1]. (And Lucas Morin created an LGBM 792 also [here][2] but we don't need two LGBM in the list)\n\n[1]: https://www.kaggle.com/code/ambrosm/amex-lightgbm-quickstart\n[2]: https://www.kaggle.com/code/lucasmorin/amex-lgbm-features-eng",
      "votes": null
    },
    {
      "id": "1810760",
      "postDate": "06/03/2022 21:25:25",
      "content": "<p>OMG, I think I worked too late. Lightgbm is the most common model. Thank you for pointing out.</p>",
      "rawMarkdown": "OMG, I think I worked too late. Lightgbm is the most common model. Thank you for pointing out.",
      "votes": null
    },
    {
      "id": "1810825",
      "postDate": "06/03/2022 23:59:02",
      "content": "<p><a href=\"https://www.kaggle.com/jingwora1\" target=\"_blank\">@jingwora1</a> FYI, in your latest update. The author and link for LGBM are incorrect.</p>",
      "rawMarkdown": "jingwora1 FYI, in your latest update. The author and link for LGBM are incorrect.",
      "votes": null
    },
    {
      "id": "1810857",
      "postDate": "06/04/2022 01:40:28",
      "content": "<p><a href=\"https://www.kaggle.com/cdeotte\" target=\"_blank\">@cdeotte</a> Thank again. The link is corrected.</p>",
      "rawMarkdown": "cdeotte Thank again. The link is corrected.",
      "votes": null
    },
    {
      "id": "1811021",
      "postDate": "06/04/2022 07:16:12",
      "content": "<p>Thanks for aggregating this, <a href=\"https://www.kaggle.com/jingwora1\" target=\"_blank\">@jingwora1</a>! </p>",
      "rawMarkdown": "Thanks for aggregating this, @jingwora1!",
      "votes": null
    },
    {
      "id": "1811531",
      "postDate": "06/04/2022 19:23:03",
      "content": "<p>now 0.793 thanks to correctly feature engineering the date :)</p>",
      "rawMarkdown": "now 0.793 thanks to correctly feature engineering the date :)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1810098,
      "author_name": "kingychiu",
      "author_url": "",
      "post_date": "06/03/2022 09:14:34",
      "content": "<p>lightautoml one is actually an ensemble of a Logistic Regression, a LightGBM, and a Catboost.</p>\n<p>From the log:</p>\n<pre><code>Final prediction for new objects (level 0) = \n     0.11480 * (5 averaged models Lvl_0_Pipe_0_Mod_0_LinearL2) +\n     0.33581 * (5 averaged models Lvl_0_Pipe_1_Mod_0_LightGBM) +\n     0.54940 * (5 averaged models Lvl_0_Pipe_1_Mod_1_CatBoost) \n</code></pre>",
      "votes": null,
      "replies": [
        {
          "id": 1810274,
          "author_name": "jingwora1",
          "author_url": "",
          "post_date": "06/03/2022 12:07:20",
          "content": "<p>Thank you for your comment.<br>\nI agree and removed automl from the list.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1810409,
      "author_name": "cdeotte",
      "author_url": "",
      "post_date": "06/03/2022 14:34:49",
      "content": "<p>Nice list. You're missing LGBM at 792 from AmbrosM <a href=\"https://www.kaggle.com/code/ambrosm/amex-lightgbm-quickstart\" target=\"_blank\">here</a>. (And Lucas Morin created an LGBM 792 also <a href=\"https://www.kaggle.com/code/lucasmorin/amex-lgbm-features-eng\" target=\"_blank\">here</a> but we don't need two LGBM in the list)</p>",
      "votes": null,
      "replies": [
        {
          "id": 1810760,
          "author_name": "jingwora1",
          "author_url": "",
          "post_date": "06/03/2022 21:25:25",
          "content": "<p>OMG, I think I worked too late. Lightgbm is the most common model. Thank you for pointing out.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1810825,
          "author_name": "cdeotte",
          "author_url": "",
          "post_date": "06/03/2022 23:59:02",
          "content": "<p><a href=\"https://www.kaggle.com/jingwora1\" target=\"_blank\">@jingwora1</a> FYI, in your latest update. The author and link for LGBM are incorrect.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1810857,
          "author_name": "jingwora1",
          "author_url": "",
          "post_date": "06/04/2022 01:40:28",
          "content": "<p><a href=\"https://www.kaggle.com/cdeotte\" target=\"_blank\">@cdeotte</a> Thank again. The link is corrected.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1811531,
          "author_name": "lucasmorin",
          "author_url": "",
          "post_date": "06/04/2022 19:23:03",
          "content": "<p>now 0.793 thanks to correctly feature engineering the date :)</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1811021,
      "author_name": "ruchi798",
      "author_url": "",
      "post_date": "06/04/2022 07:16:12",
      "content": "<p>Thanks for aggregating this, <a href=\"https://www.kaggle.com/jingwora1\" target=\"_blank\">@jingwora1</a>! </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1810078": "Ensembling different models is one of key factors to win this competition.\nThis notebook is Top model ranking based on public score.\n\nIt could be a guideline to find area to improve score.\nI try to update this list regularly. Hope it is helpful.\n\nSo far, boosting based models beat neural network.\nAnd, most of the notebook is kindly shared by GM.\n\nCaution: Each model may have difference dataset, features and optimization level. \nLower score model may not mean that model is not performing. It may show more room for improvment.\n\n| Rank | Model | Public Score | Notebooks | \n| :---: | :---: | :---: | \n| 1 | XGBoost | 0.795  🥇 |  [RAPIDS cudf Feature Engineering + XGB][xgb2] (Jiwei Liu)   <br> [xgboost-starter-0-793][xgb1] (Chris Deotte)   | \n| 2 | Catboost | 0.794  🥇 |  [amex-catboost-rounding-trick][2.1] (Jiwei Liu)   <br>[amex-catboost-0-793][2.2] (huseyincotel)  | \n| 3▲| LightGBM | 0.793 🥇|   [amex-lgbm-features-eng][4.2] (Lucas Morin) <br> [amex-lightgbm-quickstart][4.1] (AmbrosM) <br> [AMEX Default Prediction EDA & LGBM Baseline][4.3] (Kelli Belcher)  | \n| 4▼| Py-Boost | 0.792 |  [fast-metric-and-py-boost-baseline][3] (Btbpanda)    | \n| 5 | Keras Dense | 0.790 |   [amex-keras-quickstart-1-training][5] (AmbrosM)   | \n| 6 | TensorFlow Transformer | 0.789 |   [tensorflow-transformer-0-790][6] (Chris Deotte)  | \n| 7 | TensorFlow GRU | 0.789 |  [tensorflow-gru-starter-0-790][7] (Chris Deotte)    | \n| 8 | PyTorch NN | 0.787 | [AMEX PyTorch NN: Training][PTN] (Xiang Sheng)    | \n| 9▲ | PyTorch GRU |  0.786 | [AMEX PyTorch GRU: Training][PTG] (Xiang Sheng)    | \n| 10▼ | WoE | 0.757 |  [amex-credit-score-model][9]  (Gopi Durgaprasad)    | \n| 11▼ | Random Forest | 0.715 |   [best-correlated-features-with-low-correlation][10] (Bruno Gorresen Mello)  | \n\n※ As long as it is not combined with another model, it will be in this ranking. \n※ Sorry if I miss any models. Please let me know, I will correct it.\n※ Please share your private share, I plan to add into this ranking.\n\n※ Updated: 2022-06-08\n\n[xgb1]: https://www.kaggle.com/code/cdeotte/xgboost-starter-0-793\n[xgb2]: https://www.kaggle.com/code/jiweiliu/rapids-cudf-feature-engineering-xgb\n[2.1]: https://www.kaggle.com/code/jiweiliu/amex-catboost-rounding-trick\n[2.2]: https://www.kaggle.com/code/huseyincot/amex-catboost-0-793\n[3]: https://www.kaggle.com/code/btbpanda/fast-metric-and-py-boost-baseline\n[4.1]: https://www.kaggle.com/code/ambrosm/amex-lightgbm-quickstart\n[4.2]: https://www.kaggle.com/code/lucasmorin/amex-lgbm-features-eng\n[4.3]: https://www.kaggle.com/code/kellibelcher/amex-default-prediction-eda-lgbm-baseline\n[5]: https://www.kaggle.com/code/ambrosm/amex-keras-quickstart-1-training\n[6]: https://www.kaggle.com/code/cdeotte/tensorflow-transformer-0-790\n[7]: https://www.kaggle.com/code/cdeotte/tensorflow-gru-starter-0-790\n[PTN]: https://www.kaggle.com/code/voix97/amex-pytorch-nn-training\n[PTG]: https://www.kaggle.com/code/voix97/amex-pytorch-gru-training\n[9]: https://www.kaggle.com/code/gopidurgaprasad/amex-credit-score-model\n[10]: https://www.kaggle.com/code/bgmello/best-correlated-features-with-low-correlation",
    "1810098": "lightautoml one is actually an ensemble of a Logistic Regression, a LightGBM, and a Catboost.\n\nFrom the log:\n```\nFinal prediction for new objects (level 0) = \n\t 0.11480 * (5 averaged models Lvl_0_Pipe_0_Mod_0_LinearL2) +\n\t 0.33581 * (5 averaged models Lvl_0_Pipe_1_Mod_0_LightGBM) +\n\t 0.54940 * (5 averaged models Lvl_0_Pipe_1_Mod_1_CatBoost) \n```",
    "1810274": "Thank you for your comment.\nI agree and removed automl from the list.",
    "1810409": "Nice list. You're missing LGBM at 792 from AmbrosM [here][1]. (And Lucas Morin created an LGBM 792 also [here][2] but we don't need two LGBM in the list)\n\n[1]: https://www.kaggle.com/code/ambrosm/amex-lightgbm-quickstart\n[2]: https://www.kaggle.com/code/lucasmorin/amex-lgbm-features-eng",
    "1810760": "OMG, I think I worked too late. Lightgbm is the most common model. Thank you for pointing out.",
    "1810825": "jingwora1 FYI, in your latest update. The author and link for LGBM are incorrect.",
    "1810857": "cdeotte Thank again. The link is corrected.",
    "1811021": "Thanks for aggregating this, @jingwora1!",
    "1811531": "now 0.793 thanks to correctly feature engineering the date :)"
  },
  "source": "meta"
}