{
  "id": 55326,
  "title": "Very important feature share with kaggler 0.233 lightgbm",
  "url": "/competitions/avito-demand-prediction/discussion/55326",
  "author_name": "",
  "post_date": "2018-04-25T08:08:29.374605600Z",
  "votes": 11,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Dear kaggler, this is my first time to join a competitions in kaggle.</p>\n\n<p>I use a very simple <a href=\"https://www.kaggle.com/ogrellier/python-target-encoding-for-categorical-features\">mean-encoding</a> for <strong>price</strong> and <strong>deal_probability</strong>  on following category_columns:</p>\n\n<ul>\n<li>'region', </li>\n<li>'city', </li>\n<li>'parent_category_name', </li>\n<li>'category_name','</li>\n<li>'image_top_1',</li>\n<li>'user_type',</li>\n<li>'item_seq_number',</li>\n<li>'day_of_month',</li>\n<li>'day_of_week'</li>\n</ul>\n\n<p>Get a 0.233 baseline model</p>\n\n<p><img src=\"https://www.kaggleusercontent.com/kf/3351611/eyJhbGciOiJkaXIiLCJlbmMiOiJBMTI4Q0JDLUhTMjU2In0..H8COHx_291_IdyqtpbKjmQ.a3XFOEmDN_4JD5CiuewLZVNWWv6Vlcqy7PM7G_ptwOLCIKz_HhnERiNW1_w16ohHIEEupO7LWbiWSRr-iZL1HLi0CCEW8425qPFgG6wUnijAuVE2WfWdI74kTyv9pwL2txB_UY_baL58XxUa174w8g.CC2q1C-teEqFLDTOBPyt3A/__results___files/__results___15_1.png\" alt=\"LightGBM Feature importance\"></p>\n\n<p>You can see image_top_1_deal_probability_avg is very important feature in dataset, so <strong>I guess the image feature has lot of potential</strong>.</p>\n\n<p>kernel is <a href=\"https://www.kaggle.com/classtag/lightgbm-with-mean-encode-feature-0-233\">more detail code</a></p>",
  "messages": [
    {
      "id": "319096",
      "postDate": "04/25/2018 08:08:29",
      "content": "<p>Dear kaggler, this is my first time to join a competitions in kaggle.</p>\n\n<p>I use a very simple <a href=\"https://www.kaggle.com/ogrellier/python-target-encoding-for-categorical-features\">mean-encoding</a> for <strong>price</strong> and <strong>deal_probability</strong>  on following category_columns:</p>\n\n<ul>\n<li>'region', </li>\n<li>'city', </li>\n<li>'parent_category_name', </li>\n<li>'category_name','</li>\n<li>'image_top_1',</li>\n<li>'user_type',</li>\n<li>'item_seq_number',</li>\n<li>'day_of_month',</li>\n<li>'day_of_week'</li>\n</ul>\n\n<p>Get a 0.233 baseline model</p>\n\n<p><img src=\"https://www.kaggleusercontent.com/kf/3351611/eyJhbGciOiJkaXIiLCJlbmMiOiJBMTI4Q0JDLUhTMjU2In0..H8COHx_291_IdyqtpbKjmQ.a3XFOEmDN_4JD5CiuewLZVNWWv6Vlcqy7PM7G_ptwOLCIKz_HhnERiNW1_w16ohHIEEupO7LWbiWSRr-iZL1HLi0CCEW8425qPFgG6wUnijAuVE2WfWdI74kTyv9pwL2txB_UY_baL58XxUa174w8g.CC2q1C-teEqFLDTOBPyt3A/__results___files/__results___15_1.png\" alt=\"LightGBM Feature importance\"></p>\n\n<p>You can see image_top_1_deal_probability_avg is very important feature in dataset, so <strong>I guess the image feature has lot of potential</strong>.</p>\n\n<p>kernel is <a href=\"https://www.kaggle.com/classtag/lightgbm-with-mean-encode-feature-0-233\">more detail code</a></p>",
      "rawMarkdown": "Dear kaggler, this is my first time to join a competitions in kaggle.\n\nI use a very simple [mean-encoding][1] for **price** and **deal_probability**  on following category_columns:\n\n- 'region', \n- 'city', \n- 'parent_category_name', \n- 'category_name','\n- 'image_top_1',\n- 'user_type',\n- 'item_seq_number',\n- 'day_of_month',\n- 'day_of_week'\n\nGet a 0.233 baseline model\n\n![LightGBM Feature importance][2]\n\nYou can see image_top_1_deal_probability_avg is very important feature in dataset, so **I guess the image feature has lot of potential**.\n\nkernel is [more detail code][3]\n\n\n  [1]: https://www.kaggle.com/ogrellier/python-target-encoding-for-categorical-features\n  [2]: https://www.kaggleusercontent.com/kf/3351611/eyJhbGciOiJkaXIiLCJlbmMiOiJBMTI4Q0JDLUhTMjU2In0..H8COHx_291_IdyqtpbKjmQ.a3XFOEmDN_4JD5CiuewLZVNWWv6Vlcqy7PM7G_ptwOLCIKz_HhnERiNW1_w16ohHIEEupO7LWbiWSRr-iZL1HLi0CCEW8425qPFgG6wUnijAuVE2WfWdI74kTyv9pwL2txB_UY_baL58XxUa174w8g.CC2q1C-teEqFLDTOBPyt3A/__results___files/__results___15_1.png\n  [3]: https://www.kaggle.com/classtag/lightgbm-with-mean-encode-feature-0-233",
      "votes": null
    },
    {
      "id": "322943",
      "postDate": "05/04/2018 01:31:11",
      "content": "<p><code>day_of_month</code> could lead to overfit since each value only occurs once in the entire train and test set.</p>",
      "rawMarkdown": "`day_of_month` could lead to overfit since each value only occurs once in the entire train and test set.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 322943,
      "author_name": "matthewa313",
      "author_url": "",
      "post_date": "05/04/2018 01:31:11",
      "content": "<p><code>day_of_month</code> could lead to overfit since each value only occurs once in the entire train and test set.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "319096": "Dear kaggler, this is my first time to join a competitions in kaggle.\n\nI use a very simple [mean-encoding][1] for **price** and **deal_probability**  on following category_columns:\n\n- 'region', \n- 'city', \n- 'parent_category_name', \n- 'category_name','\n- 'image_top_1',\n- 'user_type',\n- 'item_seq_number',\n- 'day_of_month',\n- 'day_of_week'\n\nGet a 0.233 baseline model\n\n![LightGBM Feature importance][2]\n\nYou can see image_top_1_deal_probability_avg is very important feature in dataset, so **I guess the image feature has lot of potential**.\n\nkernel is [more detail code][3]\n\n\n  [1]: https://www.kaggle.com/ogrellier/python-target-encoding-for-categorical-features\n  [2]: https://www.kaggleusercontent.com/kf/3351611/eyJhbGciOiJkaXIiLCJlbmMiOiJBMTI4Q0JDLUhTMjU2In0..H8COHx_291_IdyqtpbKjmQ.a3XFOEmDN_4JD5CiuewLZVNWWv6Vlcqy7PM7G_ptwOLCIKz_HhnERiNW1_w16ohHIEEupO7LWbiWSRr-iZL1HLi0CCEW8425qPFgG6wUnijAuVE2WfWdI74kTyv9pwL2txB_UY_baL58XxUa174w8g.CC2q1C-teEqFLDTOBPyt3A/__results___files/__results___15_1.png\n  [3]: https://www.kaggle.com/classtag/lightgbm-with-mean-encode-feature-0-233",
    "322943": "`day_of_month` could lead to overfit since each value only occurs once in the entire train and test set."
  },
  "source": "meta"
}