{
  "id": 116651,
  "title": "DT (xgb & lgb) and NN on Time Series",
  "url": "/competitions/favorita-grocery-sales-forecasting/discussion/116651",
  "author_name": "leric",
  "post_date": "2019-11-10T14:35:45.222000",
  "votes": 1,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I've notice there is kernel using lightgbm and xgboost that perform rather well without performing detrend and deseasonalization (or that I misunderstood their code?).  One example <a href=\"https://www.kaggle.com/ceshine/lgbm-starter\">https://www.kaggle.com/ceshine/lgbm-starter</a>. I've also seen other xgb and NN algorithm that did not perform those process in their model. So is it not necessary for these algorithm?</p>",
  "messages": [
    {
      "id": 669824,
      "postDate": "2019-11-10T14:35:45.223Z",
      "content": "<p>I've notice there is kernel using lightgbm and xgboost that perform rather well without performing detrend and deseasonalization (or that I misunderstood their code?).  One example <a href=\"https://www.kaggle.com/ceshine/lgbm-starter\">https://www.kaggle.com/ceshine/lgbm-starter</a>. I've also seen other xgb and NN algorithm that did not perform those process in their model. So is it not necessary for these algorithm?</p>",
      "rawMarkdown": "I've notice there is kernel using lightgbm and xgboost that perform rather well without performing detrend and deseasonalization (or that I misunderstood their code?).  One example https://www.kaggle.com/ceshine/lgbm-starter. I've also seen other xgb and NN algorithm that did not perform those process in their model. So is it not necessary for these algorithm?",
      "votes": 1
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "669824": "I've notice there is kernel using lightgbm and xgboost that perform rather well without performing detrend and deseasonalization (or that I misunderstood their code?).  One example https://www.kaggle.com/ceshine/lgbm-starter. I've also seen other xgb and NN algorithm that did not perform those process in their model. So is it not necessary for these algorithm?"
  }
}