{
  "id": 56421,
  "title": "Some minds",
  "url": "/competitions/talkingdata-adtracking-fraud-detection/discussion/56421",
  "author_name": "cpuwdl",
  "post_date": "2018-05-09T15:26:56.409000",
  "votes": 2,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I use 28 features and day 789 hour 0-15 train data to get 0.9816/0.9809. And 34 features and all train data to get 0.9818/0.9810. Simple average to get 0.9819/0.9812. I first get this score is 30th. The last day is 120th. The final time is 400th. I use 10 days to get it without cv. If I use 3 days to do cv, maybe I can avoid this situation. So if a 100th people open submission in last 12 hours, the 120th people will move some positions. I think ensemble submission is hard to detect and maybe some one can get 80th+. If I open my submission in last 12 hours, the 140th is also affected. I think this in 4 months ago, now it become reality. The \"Mercari Price Suggestion Challenge\" also have a good kernel released in last day, but the kernel can't run in final stage. Why some people like do this? Beat someone and upgrade another one.</p>",
  "messages": [
    {
      "id": 326337,
      "postDate": "2018-05-09T15:26:56.410Z",
      "content": "<p>I use 28 features and day 789 hour 0-15 train data to get 0.9816/0.9809. And 34 features and all train data to get 0.9818/0.9810. Simple average to get 0.9819/0.9812. I first get this score is 30th. The last day is 120th. The final time is 400th. I use 10 days to get it without cv. If I use 3 days to do cv, maybe I can avoid this situation. So if a 100th people open submission in last 12 hours, the 120th people will move some positions. I think ensemble submission is hard to detect and maybe some one can get 80th+. If I open my submission in last 12 hours, the 140th is also affected. I think this in 4 months ago, now it become reality. The \"Mercari Price Suggestion Challenge\" also have a good kernel released in last day, but the kernel can't run in final stage. Why some people like do this? Beat someone and upgrade another one.</p>",
      "rawMarkdown": "I use 28 features and day 789 hour 0-15 train data to get 0.9816/0.9809. And 34 features and all train data to get 0.9818/0.9810. Simple average to get 0.9819/0.9812. I first get this score is 30th. The last day is 120th. The final time is 400th. I use 10 days to get it without cv. If I use 3 days to do cv, maybe I can avoid this situation. So if a 100th people open submission in last 12 hours, the 120th people will move some positions. I think ensemble submission is hard to detect and maybe some one can get 80th+. If I open my submission in last 12 hours, the 140th is also affected. I think this in 4 months ago, now it become reality. The \"Mercari Price Suggestion Challenge\" also have a good kernel released in last day, but the kernel can't run in final stage. Why some people like do this? Beat someone and upgrade another one.",
      "votes": 2
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "326337": "I use 28 features and day 789 hour 0-15 train data to get 0.9816/0.9809. And 34 features and all train data to get 0.9818/0.9810. Simple average to get 0.9819/0.9812. I first get this score is 30th. The last day is 120th. The final time is 400th. I use 10 days to get it without cv. If I use 3 days to do cv, maybe I can avoid this situation. So if a 100th people open submission in last 12 hours, the 120th people will move some positions. I think ensemble submission is hard to detect and maybe some one can get 80th+. If I open my submission in last 12 hours, the 140th is also affected. I think this in 4 months ago, now it become reality. The \"Mercari Price Suggestion Challenge\" also have a good kernel released in last day, but the kernel can't run in final stage. Why some people like do this? Beat someone and upgrade another one."
  }
}