{
  "id": 56566,
  "title": "anyone using xgboost",
  "url": "/competitions/trackml-particle-identification/discussion/56566",
  "author_name": "",
  "post_date": "2018-05-11T11:25:32.377851900Z",
  "votes": 7,
  "comment_count": 3,
  "views": 0,
  "content": "<p>I am wondering if anyone using traditional ML methods like xgboost, random forest?\nHow can we formulate the ML problems for such traditional methods.</p>\n\n<p>e.g. using angles, length from neighboring hits as features for training.</p>",
  "messages": [
    {
      "id": "327363",
      "postDate": "05/11/2018 11:25:32",
      "content": "<p>I am wondering if anyone using traditional ML methods like xgboost, random forest?\nHow can we formulate the ML problems for such traditional methods.</p>\n\n<p>e.g. using angles, length from neighboring hits as features for training.</p>",
      "rawMarkdown": "I am wondering if anyone using traditional ML methods like xgboost, random forest?\nHow can we formulate the ML problems for such traditional methods.\n\ne.g. using angles, length from neighboring hits as features for training.",
      "votes": null
    },
    {
      "id": "327365",
      "postDate": "05/11/2018 11:29:44",
      "content": "<p>Hi!\nProbably, you will find useful the slides in this topic <a href=\"https://www.kaggle.com/c/trackml-particle-identification/discussion/56176\">https://www.kaggle.com/c/trackml-particle-identification/discussion/56176</a>. It is non-standard problem for classifiers and based on kNN classifier, but other classifiers can also be used.  </p>",
      "rawMarkdown": "Hi!\nProbably, you will find useful the slides in this topic https://www.kaggle.com/c/trackml-particle-identification/discussion/56176. It is non-standard problem for classifiers and based on kNN classifier, but other classifiers can also be used.",
      "votes": null
    },
    {
      "id": "327423",
      "postDate": "05/11/2018 14:16:50",
      "content": "<p>plus 1, also thought about the \"classical\" kaggle methods... </p>",
      "rawMarkdown": "plus 1, also thought about the \"classical\" kaggle methods...",
      "votes": null
    },
    {
      "id": "327503",
      "postDate": "05/11/2018 17:48:52",
      "content": "<p>IMHO, once you have proper features and targets, almost any problem can be tackled (at least in part) with \"classical\" ML algos. I have had quite a bit of success with XGBoost (in conjunction with other methods) in a few image and NLP competitions. Would be interesting to see what kinds of features it's possible to come up for this problem.</p>",
      "rawMarkdown": "IMHO, once you have proper features and targets, almost any problem can be tackled (at least in part) with \"classical\" ML algos. I have had quite a bit of success with XGBoost (in conjunction with other methods) in a few image and NLP competitions. Would be interesting to see what kinds of features it's possible to come up for this problem.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 327365,
      "author_name": "mikhailhushchyn",
      "author_url": "",
      "post_date": "05/11/2018 11:29:44",
      "content": "<p>Hi!\nProbably, you will find useful the slides in this topic <a href=\"https://www.kaggle.com/c/trackml-particle-identification/discussion/56176\">https://www.kaggle.com/c/trackml-particle-identification/discussion/56176</a>. It is non-standard problem for classifiers and based on kNN classifier, but other classifiers can also be used.  </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 327423,
      "author_name": "hireme",
      "author_url": "",
      "post_date": "05/11/2018 14:16:50",
      "content": "<p>plus 1, also thought about the \"classical\" kaggle methods... </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 327503,
      "author_name": "tunguz",
      "author_url": "",
      "post_date": "05/11/2018 17:48:52",
      "content": "<p>IMHO, once you have proper features and targets, almost any problem can be tackled (at least in part) with \"classical\" ML algos. I have had quite a bit of success with XGBoost (in conjunction with other methods) in a few image and NLP competitions. Would be interesting to see what kinds of features it's possible to come up for this problem.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "327363": "I am wondering if anyone using traditional ML methods like xgboost, random forest?\nHow can we formulate the ML problems for such traditional methods.\n\ne.g. using angles, length from neighboring hits as features for training.",
    "327365": "Hi!\nProbably, you will find useful the slides in this topic https://www.kaggle.com/c/trackml-particle-identification/discussion/56176. It is non-standard problem for classifiers and based on kNN classifier, but other classifiers can also be used.",
    "327423": "plus 1, also thought about the \"classical\" kaggle methods...",
    "327503": "IMHO, once you have proper features and targets, almost any problem can be tackled (at least in part) with \"classical\" ML algos. I have had quite a bit of success with XGBoost (in conjunction with other methods) in a few image and NLP competitions. Would be interesting to see what kinds of features it's possible to come up for this problem."
  },
  "source": "meta"
}