{
  "id": 287558,
  "title": "XGB model with windowed features",
  "url": "/competitions/g-research-crypto-forecasting/discussion/287558",
  "author_name": "Curator",
  "post_date": "2021-11-14T15:02:24.615000",
  "votes": 4,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I've created an XGBoost notebook based on Tarlan Nazarov's notebook. My notebook discarded the overlapping data from the training set and also added functionality for adding lagged observations as features for training. The model does not perform well, but hopefully it will provide a starting point for others.</p>\n<p>Link: <a href=\"https://www.kaggle.com/mingyangyi/g-research-crypto-xgb-lag\" target=\"_blank\">https://www.kaggle.com/mingyangyi/g-research-crypto-xgb-lag</a></p>",
  "messages": [
    {
      "id": 1582122,
      "postDate": "2021-11-14T15:02:24.617Z",
      "content": "<p>I've created an XGBoost notebook based on Tarlan Nazarov's notebook. My notebook discarded the overlapping data from the training set and also added functionality for adding lagged observations as features for training. The model does not perform well, but hopefully it will provide a starting point for others.</p>\n<p>Link: <a href=\"https://www.kaggle.com/mingyangyi/g-research-crypto-xgb-lag\" target=\"_blank\">https://www.kaggle.com/mingyangyi/g-research-crypto-xgb-lag</a></p>",
      "rawMarkdown": "I've created an XGBoost notebook based on Tarlan Nazarov's notebook. My notebook discarded the overlapping data from the training set and also added functionality for adding lagged observations as features for training. The model does not perform well, but hopefully it will provide a starting point for others.\n\nLink: https://www.kaggle.com/mingyangyi/g-research-crypto-xgb-lag",
      "votes": 4
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1582122": "I've created an XGBoost notebook based on Tarlan Nazarov's notebook. My notebook discarded the overlapping data from the training set and also added functionality for adding lagged observations as features for training. The model does not perform well, but hopefully it will provide a starting point for others.\n\nLink: https://www.kaggle.com/mingyangyi/g-research-crypto-xgb-lag"
  }
}