{
  "id": 1338,
  "title": "Problems with the new dataset/benchmark file",
  "url": "/competitions/benchmark-bond-trade-price-challenge/discussion/1338",
  "author_name": "Adriano Azevedo-Filho",
  "post_date": "2012-02-07T00:46:26.887000",
  "votes": 0,
  "comment_count": 0,
  "views": 1703,
  "content": "<p>The new train and test files within the .zip files have missing values codified as NaN (not a number, e.g. 0/0) instead of the more usual NA symbol used in the first dataset version. I haven't checked other files wrt this issue.</p>\r\n<p>The benchmark R source uploaded in the site assumes NA as the symbol for missing values and might not work properly with current files [Edit: in this situation it doesn't seem too be an issue, R seems to understand NaN (not a number)&nbsp; as NA (missing value)].The\r\n line &quot;library(plyr)&quot; in the source can be safely deleted as this package is not used to produce the random forest benchmark.</p>",
  "messages": [
    {
      "id": 8469,
      "postDate": "2012-02-07T00:46:26.887Z",
      "content": "<p>The new train and test files within the .zip files have missing values codified as NaN (not a number, e.g. 0/0) instead of the more usual NA symbol used in the first dataset version. I haven't checked other files wrt this issue.</p>\r\n<p>The benchmark R source uploaded in the site assumes NA as the symbol for missing values and might not work properly with current files [Edit: in this situation it doesn't seem too be an issue, R seems to understand NaN (not a number)&nbsp; as NA (missing value)].The\r\n line &quot;library(plyr)&quot; in the source can be safely deleted as this package is not used to produce the random forest benchmark.</p>"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "8469": ""
  }
}