{
  "id": 89979,
  "title": "New features?",
  "url": "/competitions/LANL-Earthquake-Prediction/discussion/89979",
  "author_name": "",
  "post_date": "2019-04-19T08:34:25.707165200Z",
  "votes": 17,
  "comment_count": 11,
  "views": 0,
  "content": "<p>There seems to be an important notice about features to predict time_to_failure on this paper (at Text S4: Statistical features)\nHave you ever tried them? I’m trying to understand now.\n<a href=\"https://agupubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1002%2F2017GL074677&amp;file=grl56367-sup-0001-supinfo.pdf\">https://agupubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1002%2F2017GL074677&amp;file=grl56367-sup-0001-supinfo.pdf</a></p>",
  "messages": [
    {
      "id": "519588",
      "postDate": "04/19/2019 08:34:25",
      "content": "<p>There seems to be an important notice about features to predict time_to_failure on this paper (at Text S4: Statistical features)\nHave you ever tried them? I’m trying to understand now.\n<a href=\"https://agupubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1002%2F2017GL074677&amp;file=grl56367-sup-0001-supinfo.pdf\">https://agupubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1002%2F2017GL074677&amp;file=grl56367-sup-0001-supinfo.pdf</a></p>",
      "rawMarkdown": "There seems to be an important notice about features to predict time_to_failure on this paper (at Text S4: Statistical features)\nHave you ever tried them? I’m trying to understand now.\nhttps://agupubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1002%2F2017GL074677&amp;file=grl56367-sup-0001-supinfo.pdf",
      "votes": null
    },
    {
      "id": "519929",
      "postDate": "04/19/2019 20:34:55",
      "content": "<p>Looks like an interesting paper to check and get inspiration from. Thanks for sharing <a href=\"/toshiharuokada\">@toshiharuokada</a>. I will let you know if I find something interesting. :)</p>",
      "rawMarkdown": "Looks like an interesting paper to check and get inspiration from. Thanks for sharing @toshiharuokada. I will let you know if I find something interesting. :)",
      "votes": null
    },
    {
      "id": "520049",
      "postDate": "04/20/2019 03:33:02",
      "content": "<p>Seems they had a different scientific approach in feature engineering. Thank for sharing.</p>",
      "rawMarkdown": "Seems they had a different scientific approach in feature engineering. Thank for sharing.",
      "votes": null
    },
    {
      "id": "520199",
      "postDate": "04/20/2019 12:01:32",
      "content": "<p>Thank you for sharing this paper <a href=\"/toshiharuokada\">@toshiharuokada</a> I'll work on this method and let you know if I find something interesting 👍 .</p>",
      "rawMarkdown": "Thank you for sharing this paper @toshiharuokada I'll work on this method and let you know if I find something interesting 👍 .",
      "votes": null
    },
    {
      "id": "520301",
      "postDate": "04/20/2019 16:45:01",
      "content": "<p>Thanks for sharing. Yes, I was also thinking hard on the unique features for this earthquake project. In the URL below - \n<a href=\"https://www.sciencedaily.com/releases/2018/12/181218093045.htm\">https://www.sciencedaily.com/releases/2018/12/181218093045.htm</a> - It talked about - \"...highly predictable sound pattern that indicates slippage and fault failure,\" said Los Alamos scientist Paul Johnson. \"We also found a precise link between the fragility of the fault and the signal's strength**, which can help us more accurately predict a mega-quake.\".  </p>",
      "rawMarkdown": "Thanks for sharing. Yes, I was also thinking hard on the unique features for this earthquake project. In the URL below - \nhttps://www.sciencedaily.com/releases/2018/12/181218093045.htm - It talked about - \"...highly predictable sound pattern that indicates slippage and fault failure,\" said Los Alamos scientist Paul Johnson. \"We also found a precise link between the fragility of the fault and the signal's strength**, which can help us more accurately predict a mega-quake.\".",
      "votes": null
    },
    {
      "id": "520921",
      "postDate": "04/22/2019 01:45:38",
      "content": "<p><a href=\"/toshiharuokada\">@toshiharuokada</a> I modified a kernel from <a href=\"https://www.kaggle.com/taqanori\">taqanori</a> applied <a href=\"https://www.kaggle.com/hsinwenchang/mfcc-randomforestregressor-catboostregressor?scriptVersionId=13214884\">RandomForestRegressor &amp; CatBoostRegressor</a>, hope this kernel will help us further :)!</p>",
      "rawMarkdown": "toshiharuokada I modified a kernel from [taqanori](https://www.kaggle.com/taqanori) applied [RandomForestRegressor &amp; CatBoostRegressor](https://www.kaggle.com/hsinwenchang/mfcc-randomforestregressor-catboostregressor?scriptVersionId=13214884), hope this kernel will help us further :)!",
      "votes": null
    },
    {
      "id": "521628",
      "postDate": "04/23/2019 06:55:56",
      "content": "<p><a href=\"/hsinwenchang\">@hsinwenchang</a> Thanks for sharing your kernel. I’m stacking on XGBClassifier with Key_Error now...</p>",
      "rawMarkdown": "hsinwenchang Thanks for sharing your kernel. I’m stacking on XGBClassifier with Key_Error now...",
      "votes": null
    },
    {
      "id": "521664",
      "postDate": "04/23/2019 08:17:25",
      "content": "<p><a href=\"/toshiharuokada\">@toshiharuokada</a> you can imagine in a way that we are predict time_to_failure by regression. You can play with:\n from xgboost import XGBRegressor\nreport_cv(XGBRegressor(random_state=random_seed))\nAnd it's you share this excellent paper inspire me to modified this kernel the honor belongs to you:)!</p>",
      "rawMarkdown": "toshiharuokada you can imagine in a way that we are predict time_to_failure by regression. You can play with:\n from xgboost import XGBRegressor\nreport_cv(XGBRegressor(random_state=random_seed))\nAnd it's you share this excellent paper inspire me to modified this kernel the honor belongs to you:)!",
      "votes": null
    },
    {
      "id": "521728",
      "postDate": "04/23/2019 10:42:55",
      "content": "<p>Thanks a lot. But I want to try Classification model. I continue to hang in.</p>\n\n<h1>I’m on the way learning ML and also need to learn about classifier.</h1>",
      "rawMarkdown": "Thanks a lot. But I want to try Classification model. I continue to hang in.\n# I’m on the way learning ML and also need to learn about classifier.",
      "votes": null
    },
    {
      "id": "521791",
      "postDate": "04/23/2019 12:27:43",
      "content": "<p>Hi, you don't need to shout here (using bold and large font).</p>\n\n<p>I guess it is why you got a down vote.</p>\n\n<p>You say you want to learn.  I suggest you start by reading what people write you and try to understand it.</p>\n\n<p>Back to your comment content, what are you classifying?  Classification means a finite set of target values, typically 2.  Here the target (time to failure) is continuous, which is why we all use regression models.</p>",
      "rawMarkdown": "Hi, you don't need to shout here (using bold and large font).\n\nI guess it is why you got a down vote.\n\nYou say you want to learn.  I suggest you start by reading what people write you and try to understand it.\n\nBack to your comment content, what are you classifying?  Classification means a finite set of target values, typically 2.  Here the target (time to failure) is continuous, which is why we all use regression models.",
      "votes": null
    },
    {
      "id": "522142",
      "postDate": "04/23/2019 22:53:35",
      "content": "<p>Sorry, simply I did not know sharp means #BOLD. Intended hash tag on twitter orz.</p>",
      "rawMarkdown": "Sorry, simply I did not know sharp means #BOLD. Intended hash tag on twitter orz.",
      "votes": null
    },
    {
      "id": "522178",
      "postDate": "04/24/2019 01:04:53",
      "content": "<p>Actually, time to failure is continuous numeric value. But that is used as a name of 150,000 acoustic signal. That is why I feel something different for regressor.</p>",
      "rawMarkdown": "Actually, time to failure is continuous numeric value. But that is used as a name of 150,000 acoustic signal. That is why I feel something different for regressor.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 519929,
      "author_name": "yassinealouini",
      "author_url": "",
      "post_date": "04/19/2019 20:34:55",
      "content": "<p>Looks like an interesting paper to check and get inspiration from. Thanks for sharing <a href=\"/toshiharuokada\">@toshiharuokada</a>. I will let you know if I find something interesting. :)</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 520049,
      "author_name": "arashnic",
      "author_url": "",
      "post_date": "04/20/2019 03:33:02",
      "content": "<p>Seems they had a different scientific approach in feature engineering. Thank for sharing.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 520199,
      "author_name": "hsinwenchang",
      "author_url": "",
      "post_date": "04/20/2019 12:01:32",
      "content": "<p>Thank you for sharing this paper <a href=\"/toshiharuokada\">@toshiharuokada</a> I'll work on this method and let you know if I find something interesting 👍 .</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 520301,
      "author_name": "azc2019",
      "author_url": "",
      "post_date": "04/20/2019 16:45:01",
      "content": "<p>Thanks for sharing. Yes, I was also thinking hard on the unique features for this earthquake project. In the URL below - \n<a href=\"https://www.sciencedaily.com/releases/2018/12/181218093045.htm\">https://www.sciencedaily.com/releases/2018/12/181218093045.htm</a> - It talked about - \"...highly predictable sound pattern that indicates slippage and fault failure,\" said Los Alamos scientist Paul Johnson. \"We also found a precise link between the fragility of the fault and the signal's strength**, which can help us more accurately predict a mega-quake.\".  </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 520921,
      "author_name": "hsinwenchang",
      "author_url": "",
      "post_date": "04/22/2019 01:45:38",
      "content": "<p><a href=\"/toshiharuokada\">@toshiharuokada</a> I modified a kernel from <a href=\"https://www.kaggle.com/taqanori\">taqanori</a> applied <a href=\"https://www.kaggle.com/hsinwenchang/mfcc-randomforestregressor-catboostregressor?scriptVersionId=13214884\">RandomForestRegressor &amp; CatBoostRegressor</a>, hope this kernel will help us further :)!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 521628,
      "author_name": "toshiharuokada",
      "author_url": "",
      "post_date": "04/23/2019 06:55:56",
      "content": "<p><a href=\"/hsinwenchang\">@hsinwenchang</a> Thanks for sharing your kernel. I’m stacking on XGBClassifier with Key_Error now...</p>",
      "votes": null,
      "replies": [
        {
          "id": 521664,
          "author_name": "hsinwenchang",
          "author_url": "",
          "post_date": "04/23/2019 08:17:25",
          "content": "<p><a href=\"/toshiharuokada\">@toshiharuokada</a> you can imagine in a way that we are predict time_to_failure by regression. You can play with:\n from xgboost import XGBRegressor\nreport_cv(XGBRegressor(random_state=random_seed))\nAnd it's you share this excellent paper inspire me to modified this kernel the honor belongs to you:)!</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 521728,
          "author_name": "toshiharuokada",
          "author_url": "",
          "post_date": "04/23/2019 10:42:55",
          "content": "<p>Thanks a lot. But I want to try Classification model. I continue to hang in.</p>\n\n<h1>I’m on the way learning ML and also need to learn about classifier.</h1>",
          "votes": null,
          "replies": []
        },
        {
          "id": 521791,
          "author_name": "cpmpml",
          "author_url": "",
          "post_date": "04/23/2019 12:27:43",
          "content": "<p>Hi, you don't need to shout here (using bold and large font).</p>\n\n<p>I guess it is why you got a down vote.</p>\n\n<p>You say you want to learn.  I suggest you start by reading what people write you and try to understand it.</p>\n\n<p>Back to your comment content, what are you classifying?  Classification means a finite set of target values, typically 2.  Here the target (time to failure) is continuous, which is why we all use regression models.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 522142,
          "author_name": "toshiharuokada",
          "author_url": "",
          "post_date": "04/23/2019 22:53:35",
          "content": "<p>Sorry, simply I did not know sharp means #BOLD. Intended hash tag on twitter orz.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 522178,
      "author_name": "toshiharuokada",
      "author_url": "",
      "post_date": "04/24/2019 01:04:53",
      "content": "<p>Actually, time to failure is continuous numeric value. But that is used as a name of 150,000 acoustic signal. That is why I feel something different for regressor.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "519588": "There seems to be an important notice about features to predict time_to_failure on this paper (at Text S4: Statistical features)\nHave you ever tried them? I’m trying to understand now.\nhttps://agupubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1002%2F2017GL074677&amp;file=grl56367-sup-0001-supinfo.pdf",
    "519929": "Looks like an interesting paper to check and get inspiration from. Thanks for sharing @toshiharuokada. I will let you know if I find something interesting. :)",
    "520049": "Seems they had a different scientific approach in feature engineering. Thank for sharing.",
    "520199": "Thank you for sharing this paper @toshiharuokada I'll work on this method and let you know if I find something interesting 👍 .",
    "520301": "Thanks for sharing. Yes, I was also thinking hard on the unique features for this earthquake project. In the URL below - \nhttps://www.sciencedaily.com/releases/2018/12/181218093045.htm - It talked about - \"...highly predictable sound pattern that indicates slippage and fault failure,\" said Los Alamos scientist Paul Johnson. \"We also found a precise link between the fragility of the fault and the signal's strength**, which can help us more accurately predict a mega-quake.\".",
    "520921": "toshiharuokada I modified a kernel from [taqanori](https://www.kaggle.com/taqanori) applied [RandomForestRegressor &amp; CatBoostRegressor](https://www.kaggle.com/hsinwenchang/mfcc-randomforestregressor-catboostregressor?scriptVersionId=13214884), hope this kernel will help us further :)!",
    "521628": "hsinwenchang Thanks for sharing your kernel. I’m stacking on XGBClassifier with Key_Error now...",
    "521664": "toshiharuokada you can imagine in a way that we are predict time_to_failure by regression. You can play with:\n from xgboost import XGBRegressor\nreport_cv(XGBRegressor(random_state=random_seed))\nAnd it's you share this excellent paper inspire me to modified this kernel the honor belongs to you:)!",
    "521728": "Thanks a lot. But I want to try Classification model. I continue to hang in.\n# I’m on the way learning ML and also need to learn about classifier.",
    "521791": "Hi, you don't need to shout here (using bold and large font).\n\nI guess it is why you got a down vote.\n\nYou say you want to learn.  I suggest you start by reading what people write you and try to understand it.\n\nBack to your comment content, what are you classifying?  Classification means a finite set of target values, typically 2.  Here the target (time to failure) is continuous, which is why we all use regression models.",
    "522142": "Sorry, simply I did not know sharp means #BOLD. Intended hash tag on twitter orz.",
    "522178": "Actually, time to failure is continuous numeric value. But that is used as a name of 150,000 acoustic signal. That is why I feel something different for regressor."
  },
  "source": "meta"
}