{
  "id": 5840,
  "title": "Newbie help: Framing the problem as a classification problem",
  "url": "/competitions/accelerometer-biometric-competition/discussion/5840",
  "author_name": "",
  "post_date": "2013-09-22T08:43:47.860Z",
  "votes": null,
  "comment_count": 1,
  "views": 1930,
  "content": "<p>I am a newbie and this is my first serious attempt at a kaggle problem. After trying the kNN approach (benchmark mentioned) and a time profile combined with kNN approach I could just attain a .75ish AUC.</p>\n<p>So I am thinking of reframing the problem as a classification problem.</p>\n<p>Break the input data into 300 point sequences. Each sequence will have a device Id (label). I calculate n-number of features from those 300 points.</p>\n<p>So I have a multi-way classification input sample. (x number of points with labels which take 1 of 387 labels). I train a multi-way classification algorithm on it (say 387-way logistic regression). While predicting, I calculate the same n-features from the 300 pt test sequence and apply my 387-way model to get a map of 387 probabilities, which will be my prediction.</p>\n<p>&nbsp;</p>\n<p>Is this how generally problems like this are framed ? Do the experienced kagglers, think that this is a good framing for this problem and similar problems ?</p>\n<p>&nbsp;</p>\n<p>&nbsp;</p>\n<p>&nbsp;</p>\n<p>&nbsp;</p>\n<p>&nbsp;</p>",
  "messages": [
    {
      "id": "31333",
      "postDate": "09/22/2013 08:43:47",
      "content": "<p>I am a newbie and this is my first serious attempt at a kaggle problem. After trying the kNN approach (benchmark mentioned) and a time profile combined with kNN approach I could just attain a .75ish AUC.</p>\n<p>So I am thinking of reframing the problem as a classification problem.</p>\n<p>Break the input data into 300 point sequences. Each sequence will have a device Id (label). I calculate n-number of features from those 300 points.</p>\n<p>So I have a multi-way classification input sample. (x number of points with labels which take 1 of 387 labels). I train a multi-way classification algorithm on it (say 387-way logistic regression). While predicting, I calculate the same n-features from the 300 pt test sequence and apply my 387-way model to get a map of 387 probabilities, which will be my prediction.</p>\n<p>&nbsp;</p>\n<p>Is this how generally problems like this are framed ? Do the experienced kagglers, think that this is a good framing for this problem and similar problems ?</p>\n<p>&nbsp;</p>\n<p>&nbsp;</p>\n<p>&nbsp;</p>\n<p>&nbsp;</p>\n<p>&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "32311",
      "postDate": "10/13/2013 16:58:39",
      "content": "<p>&nbsp;</p>\n<p>Siddharth,</p>\n<p>This is very similar to how I have framed the problem.&nbsp; However, I have not generated such good results so far.&nbsp; After taking some time off, I have come back to re-visit my feature generation process to see how it can be improved.&nbsp;</p>\n<p>As this contest has progressed, more and more information has come out about all the &quot;data leaks&quot; in the contest.&nbsp; (hints about the right answer in the data without actually solving the problem the sponsor wanted to solve).&nbsp;&nbsp;&nbsp; This is why the Leaderboard AUC scores are so high.</p>\n<p>My model has not taken advantage of any of these leaks.&nbsp;</p>\n<p>Also, some of the other posted freebie solutions I have seen don't use ML. They are more probability generators.</p>\n<p>Good Luck.</p>\n<p>&nbsp;</p>\n<p>&nbsp;</p>\n<p>&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 32311,
      "author_name": "geringer",
      "author_url": "",
      "post_date": "10/13/2013 16:58:39",
      "content": "<p>&nbsp;</p>\n<p>Siddharth,</p>\n<p>This is very similar to how I have framed the problem.&nbsp; However, I have not generated such good results so far.&nbsp; After taking some time off, I have come back to re-visit my feature generation process to see how it can be improved.&nbsp;</p>\n<p>As this contest has progressed, more and more information has come out about all the &quot;data leaks&quot; in the contest.&nbsp; (hints about the right answer in the data without actually solving the problem the sponsor wanted to solve).&nbsp;&nbsp;&nbsp; This is why the Leaderboard AUC scores are so high.</p>\n<p>My model has not taken advantage of any of these leaks.&nbsp;</p>\n<p>Also, some of the other posted freebie solutions I have seen don't use ML. They are more probability generators.</p>\n<p>Good Luck.</p>\n<p>&nbsp;</p>\n<p>&nbsp;</p>\n<p>&nbsp;</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "31333": "",
    "32311": ""
  },
  "source": "meta"
}