{
  "id": 65790,
  "title": "I begin. How to start well ?",
  "url": "/competitions/ga-customer-revenue-prediction/discussion/65790",
  "author_name": "",
  "post_date": "2018-09-14T23:58:53.685636Z",
  "votes": null,
  "comment_count": 7,
  "views": 0,
  "content": "<p>Hi everybody,</p>\n\n<p>After some courses at coursera.org I decide to be more practical so I decided to participate at many Kaggle competitions.</p>\n\n<p>Where could I start ?\nShould I join a team ?  How to integrate a team ? Are there team who help beginners ?</p>\n\n<p>I would appreciate any help.\nThanks</p>",
  "messages": [
    {
      "id": "387448",
      "postDate": "09/14/2018 23:58:53",
      "content": "<p>Hi everybody,</p>\n\n<p>After some courses at coursera.org I decide to be more practical so I decided to participate at many Kaggle competitions.</p>\n\n<p>Where could I start ?\nShould I join a team ?  How to integrate a team ? Are there team who help beginners ?</p>\n\n<p>I would appreciate any help.\nThanks</p>",
      "rawMarkdown": "Hi everybody,\n\nAfter some courses at coursera.org I decide to be more practical so I decided to participate at many Kaggle competitions.\n\nWhere could I start ?\nShould I join a team ?  How to integrate a team ? Are there team who help beginners ?\n\nI would appreciate any help.\nThanks",
      "votes": null
    },
    {
      "id": "387482",
      "postDate": "09/15/2018 03:43:17",
      "content": "<p>It's easier (and much faster) to start with knowledge competitions like <a href=\"https://www.kaggle.com/c/titanic\">titanic</a> or simple datasets like <a href=\"https://www.kaggle.com/blastchar/telco-customer-churn\">Telco customer churn</a>. But if you want a real competition I'd suggest to start looking at exploratory analysis and understanding the data and target. </p>\n\n<p>After that you can try a few different models like logistic regression, random forests and gbdt.</p>",
      "rawMarkdown": "It's easier (and much faster) to start with knowledge competitions like [titanic][1] or simple datasets like [Telco customer churn][2]. But if you want a real competition I'd suggest to start looking at exploratory analysis and understanding the data and target. \n\nAfter that you can try a few different models like logistic regression, random forests and gbdt.\n\n\n  [1]: https://www.kaggle.com/c/titanic\n  [2]: https://www.kaggle.com/blastchar/telco-customer-churn",
      "votes": null
    },
    {
      "id": "387495",
      "postDate": "09/15/2018 04:19:56",
      "content": "<p>To start well you should </p>\n\n<p>1) set up you CV (lb is like just another fold of your CV) and it is very important, without this all other efforts are almost useless!</p>\n\n<p>2) start exploring data (plotting quantities, exploring feature relationship between target and features, ...)</p>\n\n<p>3) start with good models (usually people here in Kaggle start with fast  implementations of Boosting and artificial neural networks like LightGBM or ANN in Keras/Pytorch)</p>\n\n<p>4) try different feature engineering methods, this is also very important;  it is all about FE!</p>\n\n<p>5) .....</p>",
      "rawMarkdown": "To start well you should \n\n1) set up you CV (lb is like just another fold of your CV) and it is very important, without this all other efforts are almost useless!\n\n2) start exploring data (plotting quantities, exploring feature relationship between target and features, ...)\n\n3) start with good models (usually people here in Kaggle start with fast  implementations of Boosting and artificial neural networks like LightGBM or ANN in Keras/Pytorch)\n\n4) try different feature engineering methods, this is also very important;  it is all about FE!\n\n5) .....",
      "votes": null
    },
    {
      "id": "387529",
      "postDate": "09/15/2018 05:33:27",
      "content": "<p>Please where do I set up my CV ?\nI didn't see a link to do it from my profile.</p>",
      "rawMarkdown": "Please where do I set up my CV ?\nI didn't see a link to do it from my profile.",
      "votes": null
    },
    {
      "id": "387530",
      "postDate": "09/15/2018 05:34:41",
      "content": "<p>Thanks. \nI will follow your advice about real competition.</p>",
      "rawMarkdown": "Thanks. \nI will follow your advice about real competition.",
      "votes": null
    },
    {
      "id": "387531",
      "postDate": "09/15/2018 05:51:11",
      "content": "<p>CV = cross validation;  google it please: n-fold cross validation</p>",
      "rawMarkdown": "CV = cross validation;  google it please: n-fold cross validation",
      "votes": null
    },
    {
      "id": "387679",
      "postDate": "09/15/2018 12:11:36",
      "content": "<p>I though you was talking about Curriculum Vitae</p>",
      "rawMarkdown": "I though you was talking about Curriculum Vitae",
      "votes": null
    },
    {
      "id": "387914",
      "postDate": "09/15/2018 22:43:44",
      "content": "<p>I would suggest to get the best scoring kernel and use your creativity to add new features</p>",
      "rawMarkdown": "I would suggest to get the best scoring kernel and use your creativity to add new features",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 387482,
      "author_name": "jsaguiar",
      "author_url": "",
      "post_date": "09/15/2018 03:43:17",
      "content": "<p>It's easier (and much faster) to start with knowledge competitions like <a href=\"https://www.kaggle.com/c/titanic\">titanic</a> or simple datasets like <a href=\"https://www.kaggle.com/blastchar/telco-customer-churn\">Telco customer churn</a>. But if you want a real competition I'd suggest to start looking at exploratory analysis and understanding the data and target. </p>\n\n<p>After that you can try a few different models like logistic regression, random forests and gbdt.</p>",
      "votes": null,
      "replies": [
        {
          "id": 387530,
          "author_name": "maelfosso",
          "author_url": "",
          "post_date": "09/15/2018 05:34:41",
          "content": "<p>Thanks. \nI will follow your advice about real competition.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 387495,
      "author_name": "kainsama",
      "author_url": "",
      "post_date": "09/15/2018 04:19:56",
      "content": "<p>To start well you should </p>\n\n<p>1) set up you CV (lb is like just another fold of your CV) and it is very important, without this all other efforts are almost useless!</p>\n\n<p>2) start exploring data (plotting quantities, exploring feature relationship between target and features, ...)</p>\n\n<p>3) start with good models (usually people here in Kaggle start with fast  implementations of Boosting and artificial neural networks like LightGBM or ANN in Keras/Pytorch)</p>\n\n<p>4) try different feature engineering methods, this is also very important;  it is all about FE!</p>\n\n<p>5) .....</p>",
      "votes": null,
      "replies": [
        {
          "id": 387529,
          "author_name": "maelfosso",
          "author_url": "",
          "post_date": "09/15/2018 05:33:27",
          "content": "<p>Please where do I set up my CV ?\nI didn't see a link to do it from my profile.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 387531,
          "author_name": "kainsama",
          "author_url": "",
          "post_date": "09/15/2018 05:51:11",
          "content": "<p>CV = cross validation;  google it please: n-fold cross validation</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 387679,
          "author_name": "maelfosso",
          "author_url": "",
          "post_date": "09/15/2018 12:11:36",
          "content": "<p>I though you was talking about Curriculum Vitae</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 387914,
      "author_name": "mtinti",
      "author_url": "",
      "post_date": "09/15/2018 22:43:44",
      "content": "<p>I would suggest to get the best scoring kernel and use your creativity to add new features</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "387448": "Hi everybody,\n\nAfter some courses at coursera.org I decide to be more practical so I decided to participate at many Kaggle competitions.\n\nWhere could I start ?\nShould I join a team ?  How to integrate a team ? Are there team who help beginners ?\n\nI would appreciate any help.\nThanks",
    "387482": "It's easier (and much faster) to start with knowledge competitions like [titanic][1] or simple datasets like [Telco customer churn][2]. But if you want a real competition I'd suggest to start looking at exploratory analysis and understanding the data and target. \n\nAfter that you can try a few different models like logistic regression, random forests and gbdt.\n\n\n  [1]: https://www.kaggle.com/c/titanic\n  [2]: https://www.kaggle.com/blastchar/telco-customer-churn",
    "387495": "To start well you should \n\n1) set up you CV (lb is like just another fold of your CV) and it is very important, without this all other efforts are almost useless!\n\n2) start exploring data (plotting quantities, exploring feature relationship between target and features, ...)\n\n3) start with good models (usually people here in Kaggle start with fast  implementations of Boosting and artificial neural networks like LightGBM or ANN in Keras/Pytorch)\n\n4) try different feature engineering methods, this is also very important;  it is all about FE!\n\n5) .....",
    "387529": "Please where do I set up my CV ?\nI didn't see a link to do it from my profile.",
    "387530": "Thanks. \nI will follow your advice about real competition.",
    "387531": "CV = cross validation;  google it please: n-fold cross validation",
    "387679": "I though you was talking about Curriculum Vitae",
    "387914": "I would suggest to get the best scoring kernel and use your creativity to add new features"
  },
  "source": "meta"
}