{
  "id": 5165,
  "title": "Suggestions on how to analyse the Allstate Dataset",
  "url": "/competitions/ClaimPredictionChallenge/discussion/5165",
  "author_name": "Blackpanda",
  "post_date": "2013-07-21T22:31:19.863000",
  "votes": 0,
  "comment_count": 0,
  "views": 3608,
  "content": "<p>Hi,</p>\n<p>I am looking at the Allstate dataset as part of a Statistics project for College. I'd be really grateful if anyone on the final leaderboard could either answer the following questions or suggest variables/techniques on which I should concentrate.</p>\n<p>Firstly, did anyone use any imputation methods to fill in the missing data values (4 Categorical variables&nbsp;have a lot of missing values). If so what did you use? I've tried a number of methods but am running into memory issues. Did anyone break the data up and impute the subsets?</p>\n<p>Second, how did you deal with the unbalanced nature of the dataset. Did anyone use Smoote, SmooteBoost, or Rusboost? I used Rusboost but my results are not great. What method did you use?</p>\n<p>Third, I haven't used Model or Submodel in any of my initial prediction variables. Did anyone find that these were important?</p>\n<p>&nbsp;</p>\n<p>Really hope someone replies to this. I'd really appreciate some feedback.</p>\n<p>&nbsp;</p>\n<p>Regards,</p>\n<p>Darragh</p>\n<p>&nbsp;</p>",
  "messages": [
    {
      "id": 27496,
      "postDate": "2013-07-21T22:31:19.863Z",
      "content": "<p>Hi,</p>\n<p>I am looking at the Allstate dataset as part of a Statistics project for College. I'd be really grateful if anyone on the final leaderboard could either answer the following questions or suggest variables/techniques on which I should concentrate.</p>\n<p>Firstly, did anyone use any imputation methods to fill in the missing data values (4 Categorical variables&nbsp;have a lot of missing values). If so what did you use? I've tried a number of methods but am running into memory issues. Did anyone break the data up and impute the subsets?</p>\n<p>Second, how did you deal with the unbalanced nature of the dataset. Did anyone use Smoote, SmooteBoost, or Rusboost? I used Rusboost but my results are not great. What method did you use?</p>\n<p>Third, I haven't used Model or Submodel in any of my initial prediction variables. Did anyone find that these were important?</p>\n<p>&nbsp;</p>\n<p>Really hope someone replies to this. I'd really appreciate some feedback.</p>\n<p>&nbsp;</p>\n<p>Regards,</p>\n<p>Darragh</p>\n<p>&nbsp;</p>"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "27496": ""
  }
}