{
  "id": 53225,
  "title": "Best single model public lb score?",
  "url": "/competitions/talkingdata-adtracking-fraud-detection/discussion/53225",
  "author_name": "",
  "post_date": "2018-03-28T10:54:10.040761800Z",
  "votes": null,
  "comment_count": 10,
  "views": 0,
  "content": "<p>Can someone tell what's there best single model public leaderboard score?</p>",
  "messages": [
    {
      "id": "305007",
      "postDate": "03/28/2018 10:54:10",
      "content": "<p>Can someone tell what's there best single model public leaderboard score?</p>",
      "rawMarkdown": "Can someone tell what's there best single model public leaderboard score?",
      "votes": null
    },
    {
      "id": "305036",
      "postDate": "03/28/2018 11:48:45",
      "content": "<p>edit:  I misunderstood the question.  </p>",
      "rawMarkdown": "edit:  I misunderstood the question.",
      "votes": null
    },
    {
      "id": "305043",
      "postDate": "03/28/2018 12:23:19",
      "content": "<p>i meant what auc score you got by using a single model i.e. no blending.</p>",
      "rawMarkdown": "i meant what auc score you got by using a single model i.e. no blending.",
      "votes": null
    },
    {
      "id": "305100",
      "postDate": "03/28/2018 13:37:20",
      "content": "<p>Some numbers are given in this topic: <a href=\"https://www.kaggle.com/c/talkingdata-adtracking-fraud-detection/discussion/53141\">https://www.kaggle.com/c/talkingdata-adtracking-fraud-detection/discussion/53141</a></p>",
      "rawMarkdown": "Some numbers are given in this topic: https://www.kaggle.com/c/talkingdata-adtracking-fraud-detection/discussion/53141",
      "votes": null
    },
    {
      "id": "305108",
      "postDate": "03/28/2018 13:49:29",
      "content": "<p>best single model auc on public lb is 0.9714,  my estimation of private lb auc of this model is 0.9765. Long way to go to catch up with top rankers....There must be some magic feature that could give about 0.01 improvement on public lb, top rankers jump from 0.97 to 0.98 very fast</p>",
      "rawMarkdown": "best single model auc on public lb is 0.9714,  my estimation of private lb auc of this model is 0.9765. Long way to go to catch up with top rankers....There must be some magic feature that could give about 0.01 improvement on public lb, top rankers jump from 0.97 to 0.98 very fast",
      "votes": null
    },
    {
      "id": "305117",
      "postDate": "03/28/2018 14:02:02",
      "content": "<p>I am still at <code>0.9690</code> with chunk of max 75 million observations. Yet to try for full dataset. \nI remember one of the top guy posted few days back that his current best submission is <code>0.9764</code> with a simple model with 4 custom features. </p>",
      "rawMarkdown": "I am still at `0.9690` with chunk of max 75 million observations. Yet to try for full dataset. \nI remember one of the top guy posted few days back that his current best submission is `0.9764` with a simple model with 4 custom features.",
      "votes": null
    },
    {
      "id": "305137",
      "postDate": "03/28/2018 14:21:20",
      "content": "<p>Can I ask how many rows do you use for training?</p>",
      "rawMarkdown": "Can I ask how many rows do you use for training?",
      "votes": null
    },
    {
      "id": "305141",
      "postDate": "03/28/2018 14:22:59",
      "content": "<p>all</p>",
      "rawMarkdown": "all",
      "votes": null
    },
    {
      "id": "305228",
      "postDate": "03/28/2018 16:03:42",
      "content": "<p>Thanks for the pointer.</p>",
      "rawMarkdown": "Thanks for the pointer.",
      "votes": null
    },
    {
      "id": "307583",
      "postDate": "04/02/2018 04:22:43",
      "content": "<p>Currently our best single model achieved 0.9759 in Public LB. Haven't started ensemble yet.</p>",
      "rawMarkdown": "Currently our best single model achieved 0.9759 in Public LB. Haven't started ensemble yet.",
      "votes": null
    },
    {
      "id": "307635",
      "postDate": "04/02/2018 06:29:00",
      "content": "<p>Thanks for sharing! I would guess most of the top50 are still single models.</p>",
      "rawMarkdown": "Thanks for sharing! I would guess most of the top50 are still single models.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 305036,
      "author_name": "cpmpml",
      "author_url": "",
      "post_date": "03/28/2018 11:48:45",
      "content": "<p>edit:  I misunderstood the question.  </p>",
      "votes": null,
      "replies": [
        {
          "id": 305043,
          "author_name": "nitishsingh41",
          "author_url": "",
          "post_date": "03/28/2018 12:23:19",
          "content": "<p>i meant what auc score you got by using a single model i.e. no blending.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 305100,
          "author_name": "cpmpml",
          "author_url": "",
          "post_date": "03/28/2018 13:37:20",
          "content": "<p>Some numbers are given in this topic: <a href=\"https://www.kaggle.com/c/talkingdata-adtracking-fraud-detection/discussion/53141\">https://www.kaggle.com/c/talkingdata-adtracking-fraud-detection/discussion/53141</a></p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 305228,
          "author_name": "yipcma",
          "author_url": "",
          "post_date": "03/28/2018 16:03:42",
          "content": "<p>Thanks for the pointer.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 305108,
      "author_name": "wythhh",
      "author_url": "",
      "post_date": "03/28/2018 13:49:29",
      "content": "<p>best single model auc on public lb is 0.9714,  my estimation of private lb auc of this model is 0.9765. Long way to go to catch up with top rankers....There must be some magic feature that could give about 0.01 improvement on public lb, top rankers jump from 0.97 to 0.98 very fast</p>",
      "votes": null,
      "replies": [
        {
          "id": 305137,
          "author_name": "fatihozturk",
          "author_url": "",
          "post_date": "03/28/2018 14:21:20",
          "content": "<p>Can I ask how many rows do you use for training?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 305141,
          "author_name": "wythhh",
          "author_url": "",
          "post_date": "03/28/2018 14:22:59",
          "content": "<p>all</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 305117,
      "author_name": "pranav84",
      "author_url": "",
      "post_date": "03/28/2018 14:02:02",
      "content": "<p>I am still at <code>0.9690</code> with chunk of max 75 million observations. Yet to try for full dataset. \nI remember one of the top guy posted few days back that his current best submission is <code>0.9764</code> with a simple model with 4 custom features. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 307583,
      "author_name": "chengju",
      "author_url": "",
      "post_date": "04/02/2018 04:22:43",
      "content": "<p>Currently our best single model achieved 0.9759 in Public LB. Haven't started ensemble yet.</p>",
      "votes": null,
      "replies": [
        {
          "id": 307635,
          "author_name": "asparuhhristov",
          "author_url": "",
          "post_date": "04/02/2018 06:29:00",
          "content": "<p>Thanks for sharing! I would guess most of the top50 are still single models.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "305007": "Can someone tell what's there best single model public leaderboard score?",
    "305036": "edit:  I misunderstood the question.",
    "305043": "i meant what auc score you got by using a single model i.e. no blending.",
    "305100": "Some numbers are given in this topic: https://www.kaggle.com/c/talkingdata-adtracking-fraud-detection/discussion/53141",
    "305108": "best single model auc on public lb is 0.9714,  my estimation of private lb auc of this model is 0.9765. Long way to go to catch up with top rankers....There must be some magic feature that could give about 0.01 improvement on public lb, top rankers jump from 0.97 to 0.98 very fast",
    "305117": "I am still at `0.9690` with chunk of max 75 million observations. Yet to try for full dataset. \nI remember one of the top guy posted few days back that his current best submission is `0.9764` with a simple model with 4 custom features.",
    "305137": "Can I ask how many rows do you use for training?",
    "305141": "all",
    "305228": "Thanks for the pointer.",
    "307583": "Currently our best single model achieved 0.9759 in Public LB. Haven't started ensemble yet.",
    "307635": "Thanks for sharing! I would guess most of the top50 are still single models."
  },
  "source": "meta"
}