{
  "id": 24063,
  "title": "Cross Validation Method",
  "url": "/competitions/bosch-production-line-performance/discussion/24063",
  "author_name": "",
  "post_date": "2016-10-04T02:04:45.527Z",
  "votes": null,
  "comment_count": 2,
  "views": 487,
  "content": "<p>Hi everyone, I've noticed in some of the threads that k-fold cross validation is a popular method among Kagglers.\njust to make sure, I've been trying to implement this using the <code>score</code> method from <code>RandomForestClassifier</code> how ever I get something like 0.9ish every time (This makes sense since the the training dataset has very littel data with Response=1) and my LB seems to rank pretty low. I am wonder if I am doing the k-folds cross validation right?</p>",
  "messages": [
    {
      "id": "137684",
      "postDate": "10/04/2016 02:04:45",
      "content": "<p>Hi everyone, I've noticed in some of the threads that k-fold cross validation is a popular method among Kagglers.\njust to make sure, I've been trying to implement this using the <code>score</code> method from <code>RandomForestClassifier</code> how ever I get something like 0.9ish every time (This makes sense since the the training dataset has very littel data with Response=1) and my LB seems to rank pretty low. I am wonder if I am doing the k-folds cross validation right?</p>",
      "rawMarkdown": "Hi everyone, I've noticed in some of the threads that k-fold cross validation is a popular method among Kagglers.\r\njust to make sure, I've been trying to implement this using the ```score``` method from ```RandomForestClassifier``` how ever I get something like 0.9ish every time (This makes sense since the the training dataset has very littel data with Response=1) and my LB seems to rank pretty low. I am wonder if I am doing the k-folds cross validation right?",
      "votes": null
    },
    {
      "id": "137701",
      "postDate": "10/04/2016 06:48:54",
      "content": "<p>Which score are you using? The AUC or MCC?</p>",
      "rawMarkdown": "Which score are you using? The AUC or MCC?",
      "votes": null
    },
    {
      "id": "137706",
      "postDate": "10/04/2016 07:43:32",
      "content": "<p>How much of the data are you using for your cross validation? If your only using a small part of the data in your local CV then the results between your CV and the LB will vary more! </p>",
      "rawMarkdown": "How much of the data are you using for your cross validation? If your only using a small part of the data in your local CV then the results between your CV and the LB will vary more!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 137701,
      "author_name": "",
      "author_url": "",
      "post_date": "10/04/2016 06:48:54",
      "content": "<p>Which score are you using? The AUC or MCC?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 137706,
      "author_name": "malten",
      "author_url": "",
      "post_date": "10/04/2016 07:43:32",
      "content": "<p>How much of the data are you using for your cross validation? If your only using a small part of the data in your local CV then the results between your CV and the LB will vary more! </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "137684": "Hi everyone, I've noticed in some of the threads that k-fold cross validation is a popular method among Kagglers.\r\njust to make sure, I've been trying to implement this using the ```score``` method from ```RandomForestClassifier``` how ever I get something like 0.9ish every time (This makes sense since the the training dataset has very littel data with Response=1) and my LB seems to rank pretty low. I am wonder if I am doing the k-folds cross validation right?",
    "137701": "Which score are you using? The AUC or MCC?",
    "137706": "How much of the data are you using for your cross validation? If your only using a small part of the data in your local CV then the results between your CV and the LB will vary more!"
  },
  "source": "meta"
}