{
  "id": 121137,
  "title": "Correlation with meter_readings are almost zero",
  "url": "/competitions/ashrae-energy-prediction/discussion/121137",
  "author_name": "rackson3861",
  "post_date": "2019-12-11T12:17:28.064000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I have tried catboost and xgboost with imputing the values by ffill assuming the weather does not changes by huge amount instantly. Dropped certain features which are missing in large amounts. But these models are not giving good results. \nSplitted training and testing data in time series way. First 80% training and rest20% rest testing</p>\n\n<p>I am getting R2 score of 0.8 only.</p>",
  "messages": [
    {
      "id": 692543,
      "postDate": "2019-12-11T12:17:28.063Z",
      "content": "<p>I have tried catboost and xgboost with imputing the values by ffill assuming the weather does not changes by huge amount instantly. Dropped certain features which are missing in large amounts. But these models are not giving good results. \nSplitted training and testing data in time series way. First 80% training and rest20% rest testing</p>\n\n<p>I am getting R2 score of 0.8 only.</p>",
      "rawMarkdown": "I have tried catboost and xgboost with imputing the values by ffill assuming the weather does not changes by huge amount instantly. Dropped certain features which are missing in large amounts. But these models are not giving good results. \nSplitted training and testing data in time series way. First 80% training and rest20% rest testing\n\nI am getting R2 score of 0.8 only."
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "692543": "I have tried catboost and xgboost with imputing the values by ffill assuming the weather does not changes by huge amount instantly. Dropped certain features which are missing in large amounts. But these models are not giving good results. \nSplitted training and testing data in time series way. First 80% training and rest20% rest testing\n\nI am getting R2 score of 0.8 only."
  }
}