{
  "id": 133023,
  "title": "How are your validation scores ??",
  "url": "/competitions/google-cloud-ncaa-march-madness-2020-division-1-womens-tournament/discussion/133023",
  "author_name": "",
  "post_date": "2020-02-29T09:17:28.396523100Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Let's discuss validation 🏀 </p>\n\n<p>Here are mine.</p>\n\n<p>feature : 4 features from regular season compact results\nmodel : logistic regression\nCV : group k fold by season</p>\n\n<p>Season logloss\n   1998   0.560\n   1999   0.460\n   2000   0.425\n   2001   0.486\n   2002   0.379\n   2003   0.371\n   2004   0.493\n   2005   0.460\n   2006   0.370\n   2007   0.518\n   2008   0.372\n   2009   0.412\n   2010   0.422\n   2011   0.388\n   2012   0.343\n   2013   0.465\n   2014   0.383\n   2015   0.358\n   2016   0.449\n   2017   0.402\n   2018   0.455\n   2019   0.381</p>\n\n<p>mean: 0.425\nstd: 0.057</p>",
  "messages": [
    {
      "id": "759644",
      "postDate": "02/29/2020 09:17:28",
      "content": "<p>Let's discuss validation 🏀 </p>\n\n<p>Here are mine.</p>\n\n<p>feature : 4 features from regular season compact results\nmodel : logistic regression\nCV : group k fold by season</p>\n\n<p>Season logloss\n   1998   0.560\n   1999   0.460\n   2000   0.425\n   2001   0.486\n   2002   0.379\n   2003   0.371\n   2004   0.493\n   2005   0.460\n   2006   0.370\n   2007   0.518\n   2008   0.372\n   2009   0.412\n   2010   0.422\n   2011   0.388\n   2012   0.343\n   2013   0.465\n   2014   0.383\n   2015   0.358\n   2016   0.449\n   2017   0.402\n   2018   0.455\n   2019   0.381</p>\n\n<p>mean: 0.425\nstd: 0.057</p>",
      "rawMarkdown": "Let's discuss validation 🏀 \n\nHere are mine.\n\nfeature : 4 features from regular season compact results\nmodel : logistic regression\nCV : group k fold by season\n\n   Season logloss\n   1998   0.560\n   1999   0.460\n   2000   0.425\n   2001   0.486\n   2002   0.379\n   2003   0.371\n   2004   0.493\n   2005   0.460\n   2006   0.370\n   2007   0.518\n   2008   0.372\n   2009   0.412\n   2010   0.422\n   2011   0.388\n   2012   0.343\n   2013   0.465\n   2014   0.383\n   2015   0.358\n   2016   0.449\n   2017   0.402\n   2018   0.455\n   2019   0.381\n\nmean: 0.425\nstd: 0.057",
      "votes": null
    },
    {
      "id": "759688",
      "postDate": "02/29/2020 10:40:25",
      "content": "<p>Are you sure not using information from the future with group k-fold such as predicting 2010 with results from 2011? \nSome of the scores look too good to be true. \nWith basic features, on average I got 0.49 with xgboost and will play with features to make some improvements.</p>",
      "rawMarkdown": "Are you sure not using information from the future with group k-fold such as predicting 2010 with results from 2011? \nSome of the scores look too good to be true. \nWith basic features, on average I got 0.49 with xgboost and will play with features to make some improvements.",
      "votes": null
    },
    {
      "id": "759725",
      "postDate": "02/29/2020 11:40:04",
      "content": "<p>Thanks, Efehan.\nI see the risk in using future information but I got my features from regular seasons of the each year carefully. So, I think there is no problem.</p>",
      "rawMarkdown": "Thanks, Efehan.\nI see the risk in using future information but I got my features from regular seasons of the each year carefully. So, I think there is no problem.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 759688,
      "author_name": "efehandanisman",
      "author_url": "",
      "post_date": "02/29/2020 10:40:25",
      "content": "<p>Are you sure not using information from the future with group k-fold such as predicting 2010 with results from 2011? \nSome of the scores look too good to be true. \nWith basic features, on average I got 0.49 with xgboost and will play with features to make some improvements.</p>",
      "votes": null,
      "replies": [
        {
          "id": 759725,
          "author_name": "kur0cky",
          "author_url": "",
          "post_date": "02/29/2020 11:40:04",
          "content": "<p>Thanks, Efehan.\nI see the risk in using future information but I got my features from regular seasons of the each year carefully. So, I think there is no problem.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "759644": "Let's discuss validation 🏀 \n\nHere are mine.\n\nfeature : 4 features from regular season compact results\nmodel : logistic regression\nCV : group k fold by season\n\n   Season logloss\n   1998   0.560\n   1999   0.460\n   2000   0.425\n   2001   0.486\n   2002   0.379\n   2003   0.371\n   2004   0.493\n   2005   0.460\n   2006   0.370\n   2007   0.518\n   2008   0.372\n   2009   0.412\n   2010   0.422\n   2011   0.388\n   2012   0.343\n   2013   0.465\n   2014   0.383\n   2015   0.358\n   2016   0.449\n   2017   0.402\n   2018   0.455\n   2019   0.381\n\nmean: 0.425\nstd: 0.057",
    "759688": "Are you sure not using information from the future with group k-fold such as predicting 2010 with results from 2011? \nSome of the scores look too good to be true. \nWith basic features, on average I got 0.49 with xgboost and will play with features to make some improvements.",
    "759725": "Thanks, Efehan.\nI see the risk in using future information but I got my features from regular seasons of the each year carefully. So, I think there is no problem."
  },
  "source": "meta"
}