{
  "id": 22109,
  "title": "Will this idea work with more RAM?",
  "url": "/competitions/grupo-bimbo-inventory-demand/discussion/22109",
  "author_name": "",
  "post_date": "2016-07-07T22:57:09.700Z",
  "votes": null,
  "comment_count": 6,
  "views": 1098,
  "content": "<p>The idea is to take averages like the leaderboard script does but then put the values onto our training set and use them as covariates. The best covariate is missing on around 1.4mil rows in the test set and to overcome this I just made two separate models, one with and one without it. </p>\n\n<p>You can also do this same idea with random Forests or whatever model you want. I imagine that people high up on the leaderboard are already doing this kind of feature engineering</p>\n\n<p><a href=\"https://www.kaggle.com/abednadir/grupo-bimbo-inventory-demand/throwing-in-the-towel-here-s-my-script\">https://www.kaggle.com/abednadir/grupo-bimbo-inventory-demand/throwing-in-the-towel-here-s-my-script</a></p>\n\n<p>edit: the variable &quot;feature.names&quot; currently tries to exclude a 17th variable, you may need to edit that out</p>",
  "messages": [
    {
      "id": "126367",
      "postDate": "07/07/2016 22:57:09",
      "content": "<p>The idea is to take averages like the leaderboard script does but then put the values onto our training set and use them as covariates. The best covariate is missing on around 1.4mil rows in the test set and to overcome this I just made two separate models, one with and one without it. </p>\n\n<p>You can also do this same idea with random Forests or whatever model you want. I imagine that people high up on the leaderboard are already doing this kind of feature engineering</p>\n\n<p><a href=\"https://www.kaggle.com/abednadir/grupo-bimbo-inventory-demand/throwing-in-the-towel-here-s-my-script\">https://www.kaggle.com/abednadir/grupo-bimbo-inventory-demand/throwing-in-the-towel-here-s-my-script</a></p>\n\n<p>edit: the variable &quot;feature.names&quot; currently tries to exclude a 17th variable, you may need to edit that out</p>",
      "rawMarkdown": "The idea is to take averages like the leaderboard script does but then put the values onto our training set and use them as covariates. The best covariate is missing on around 1.4mil rows in the test set and to overcome this I just made two separate models, one with and one without it. \r\n\r\nYou can also do this same idea with random Forests or whatever model you want. I imagine that people high up on the leaderboard are already doing this kind of feature engineering\r\n\r\nhttps://www.kaggle.com/abednadir/grupo-bimbo-inventory-demand/throwing-in-the-towel-here-s-my-script\r\n\r\nedit: the variable \"feature.names\" currently tries to exclude a 17th variable, you may need to edit that out",
      "votes": null
    },
    {
      "id": "126415",
      "postDate": "07/08/2016 09:04:56",
      "content": "<p>The 9th version of the h2o-gbm script (here: <a href=\"https://www.kaggle.com/mlandry/grupo-bimbo-inventory-demand/h2o-gbm/run/265187\">https://www.kaggle.com/mlandry/grupo-bimbo-inventory-demand/h2o-gbm/run/265187</a>) does exactly that:  construct client/product key performance indicators from past data and add them to train and test.  It is a natural step IMHO. </p>",
      "rawMarkdown": "The 9th version of the h2o-gbm script (here: https://www.kaggle.com/mlandry/grupo-bimbo-inventory-demand/h2o-gbm/run/265187) does exactly that:  construct client/product key performance indicators from past data and add them to train and test.  It is a natural step IMHO.",
      "votes": null
    },
    {
      "id": "126456",
      "postDate": "07/08/2016 15:21:58",
      "content": "<p>The average helps, but not that much. LOL</p>",
      "rawMarkdown": "The average helps, but not that much. LOL",
      "votes": null
    },
    {
      "id": "126485",
      "postDate": "07/08/2016 20:09:28",
      "content": "<p>[quote=Jianbo Ye;126456]</p>\n\n<p>The average helps, but not that much. LOL</p>\n\n<p>[/quote]</p>\n\n<p>would you be so kind as to point us in the right direction? I would understand if you don't want to share your secret sauce</p>",
      "rawMarkdown": "[quote=Jianbo Ye;126456]\r\n\r\nThe average helps, but not that much. LOL\r\n\r\n[/quote]\r\n\r\nwould you be so kind as to point us in the right direction? I would understand if you don't want to share your secret sauce",
      "votes": null
    },
    {
      "id": "126491",
      "postDate": "07/08/2016 20:45:44",
      "content": "<p>[quote=SteveKane;126485]</p>\n\n<p>[quote=Jianbo Ye;126456]</p>\n\n<p>The average helps, but not that much. LOL</p>\n\n<p>[/quote]</p>\n\n<p>would you be so kind as to point us in the right direction? I would understand if you don't want to share your secret sauce</p>\n\n<p>[/quote]</p>\n\n<p>I am also wondering what other leaderboard runners are doing now! This is actually my first kaggle, and I learnt from them by researching how they win other competitions. </p>",
      "rawMarkdown": "[quote=SteveKane;126485]\r\n\r\n[quote=Jianbo Ye;126456]\r\n\r\nThe average helps, but not that much. LOL\r\n\r\n[/quote]\r\n\r\nwould you be so kind as to point us in the right direction? I would understand if you don't want to share your secret sauce\r\n\r\n[/quote]\r\n\r\nI am also wondering what other leaderboard runners are doing now! This is actually my first kaggle, and I learnt from them by researching how they win other competitions.",
      "votes": null
    },
    {
      "id": "126493",
      "postDate": "07/08/2016 20:51:28",
      "content": "<p>[quote=Jianbo Ye;126491]</p>\n\n<p>[quote=SteveKane;126485]</p>\n\n<p>[quote=Jianbo Ye;126456]</p>\n\n<p>The average helps, but not that much. LOL</p>\n\n<p>[/quote]</p>\n\n<p>would you be so kind as to point us in the right direction? I would understand if you don't want to share your secret sauce</p>\n\n<p>[/quote]</p>\n\n<p>I am also wondering what other leaderboard runners are doing now! This is actually my first kaggle, and I learnt from them by researching how they win other competitions. </p>\n\n<p>[/quote]</p>\n\n<p>Why did you join 5 years ago but not participate until now :p</p>",
      "rawMarkdown": "[quote=Jianbo Ye;126491]\r\n\r\n[quote=SteveKane;126485]\r\n\r\n[quote=Jianbo Ye;126456]\r\n\r\nThe average helps, but not that much. LOL\r\n\r\n[/quote]\r\n\r\nwould you be so kind as to point us in the right direction? I would understand if you don't want to share your secret sauce\r\n\r\n[/quote]\r\n\r\nI am also wondering what other leaderboard runners are doing now! This is actually my first kaggle, and I learnt from them by researching how they win other competitions. \r\n\r\n\r\n[/quote]\r\n\r\nWhy did you join 5 years ago but not participate until now :p",
      "votes": null
    },
    {
      "id": "126494",
      "postDate": "07/08/2016 20:55:12",
      "content": "<p>[quote=SteveKane;126493]</p>\n\n<p>[quote=Jianbo Ye;126491]</p>\n\n<p>[quote=SteveKane;126485]</p>\n\n<p>[quote=Jianbo Ye;126456]</p>\n\n<p>The average helps, but not that much. LOL</p>\n\n<p>[/quote]</p>\n\n<p>would you be so kind as to point us in the right direction? I would understand if you don't want to share your secret sauce</p>\n\n<p>[/quote]</p>\n\n<p>I am also wondering what other leaderboard runners are doing now! This is actually my first kaggle, and I learnt from them by researching how they win other competitions. </p>\n\n<p>[/quote]</p>\n\n<p>Why did you join 5 years ago but not participate until now :p</p>\n\n<p>[/quote]</p>\n\n<p>Good question, I don't know either. But now I found Kaggle is too addictive, I can't help but continuously work on the problem. </p>",
      "rawMarkdown": "[quote=SteveKane;126493]\r\n\r\n[quote=Jianbo Ye;126491]\r\n\r\n[quote=SteveKane;126485]\r\n\r\n[quote=Jianbo Ye;126456]\r\n\r\nThe average helps, but not that much. LOL\r\n\r\n[/quote]\r\n\r\nwould you be so kind as to point us in the right direction? I would understand if you don't want to share your secret sauce\r\n\r\n[/quote]\r\n\r\nI am also wondering what other leaderboard runners are doing now! This is actually my first kaggle, and I learnt from them by researching how they win other competitions. \r\n\r\n\r\n[/quote]\r\n\r\nWhy did you join 5 years ago but not participate until now :p\r\n\r\n\r\n[/quote]\r\n\r\nGood question, I don't know either. But now I found Kaggle is too addictive, I can't help but continuously work on the problem.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 126415,
      "author_name": "mrooijer",
      "author_url": "",
      "post_date": "07/08/2016 09:04:56",
      "content": "<p>The 9th version of the h2o-gbm script (here: <a href=\"https://www.kaggle.com/mlandry/grupo-bimbo-inventory-demand/h2o-gbm/run/265187\">https://www.kaggle.com/mlandry/grupo-bimbo-inventory-demand/h2o-gbm/run/265187</a>) does exactly that:  construct client/product key performance indicators from past data and add them to train and test.  It is a natural step IMHO. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 126456,
      "author_name": "amazingbob",
      "author_url": "",
      "post_date": "07/08/2016 15:21:58",
      "content": "<p>The average helps, but not that much. LOL</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 126485,
      "author_name": "abednadir",
      "author_url": "",
      "post_date": "07/08/2016 20:09:28",
      "content": "<p>[quote=Jianbo Ye;126456]</p>\n\n<p>The average helps, but not that much. LOL</p>\n\n<p>[/quote]</p>\n\n<p>would you be so kind as to point us in the right direction? I would understand if you don't want to share your secret sauce</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 126491,
      "author_name": "amazingbob",
      "author_url": "",
      "post_date": "07/08/2016 20:45:44",
      "content": "<p>[quote=SteveKane;126485]</p>\n\n<p>[quote=Jianbo Ye;126456]</p>\n\n<p>The average helps, but not that much. LOL</p>\n\n<p>[/quote]</p>\n\n<p>would you be so kind as to point us in the right direction? I would understand if you don't want to share your secret sauce</p>\n\n<p>[/quote]</p>\n\n<p>I am also wondering what other leaderboard runners are doing now! This is actually my first kaggle, and I learnt from them by researching how they win other competitions. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 126493,
      "author_name": "abednadir",
      "author_url": "",
      "post_date": "07/08/2016 20:51:28",
      "content": "<p>[quote=Jianbo Ye;126491]</p>\n\n<p>[quote=SteveKane;126485]</p>\n\n<p>[quote=Jianbo Ye;126456]</p>\n\n<p>The average helps, but not that much. LOL</p>\n\n<p>[/quote]</p>\n\n<p>would you be so kind as to point us in the right direction? I would understand if you don't want to share your secret sauce</p>\n\n<p>[/quote]</p>\n\n<p>I am also wondering what other leaderboard runners are doing now! This is actually my first kaggle, and I learnt from them by researching how they win other competitions. </p>\n\n<p>[/quote]</p>\n\n<p>Why did you join 5 years ago but not participate until now :p</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 126494,
      "author_name": "amazingbob",
      "author_url": "",
      "post_date": "07/08/2016 20:55:12",
      "content": "<p>[quote=SteveKane;126493]</p>\n\n<p>[quote=Jianbo Ye;126491]</p>\n\n<p>[quote=SteveKane;126485]</p>\n\n<p>[quote=Jianbo Ye;126456]</p>\n\n<p>The average helps, but not that much. LOL</p>\n\n<p>[/quote]</p>\n\n<p>would you be so kind as to point us in the right direction? I would understand if you don't want to share your secret sauce</p>\n\n<p>[/quote]</p>\n\n<p>I am also wondering what other leaderboard runners are doing now! This is actually my first kaggle, and I learnt from them by researching how they win other competitions. </p>\n\n<p>[/quote]</p>\n\n<p>Why did you join 5 years ago but not participate until now :p</p>\n\n<p>[/quote]</p>\n\n<p>Good question, I don't know either. But now I found Kaggle is too addictive, I can't help but continuously work on the problem. </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "126367": "The idea is to take averages like the leaderboard script does but then put the values onto our training set and use them as covariates. The best covariate is missing on around 1.4mil rows in the test set and to overcome this I just made two separate models, one with and one without it. \r\n\r\nYou can also do this same idea with random Forests or whatever model you want. I imagine that people high up on the leaderboard are already doing this kind of feature engineering\r\n\r\nhttps://www.kaggle.com/abednadir/grupo-bimbo-inventory-demand/throwing-in-the-towel-here-s-my-script\r\n\r\nedit: the variable \"feature.names\" currently tries to exclude a 17th variable, you may need to edit that out",
    "126415": "The 9th version of the h2o-gbm script (here: https://www.kaggle.com/mlandry/grupo-bimbo-inventory-demand/h2o-gbm/run/265187) does exactly that:  construct client/product key performance indicators from past data and add them to train and test.  It is a natural step IMHO.",
    "126456": "The average helps, but not that much. LOL",
    "126485": "[quote=Jianbo Ye;126456]\r\n\r\nThe average helps, but not that much. LOL\r\n\r\n[/quote]\r\n\r\nwould you be so kind as to point us in the right direction? I would understand if you don't want to share your secret sauce",
    "126491": "[quote=SteveKane;126485]\r\n\r\n[quote=Jianbo Ye;126456]\r\n\r\nThe average helps, but not that much. LOL\r\n\r\n[/quote]\r\n\r\nwould you be so kind as to point us in the right direction? I would understand if you don't want to share your secret sauce\r\n\r\n[/quote]\r\n\r\nI am also wondering what other leaderboard runners are doing now! This is actually my first kaggle, and I learnt from them by researching how they win other competitions.",
    "126493": "[quote=Jianbo Ye;126491]\r\n\r\n[quote=SteveKane;126485]\r\n\r\n[quote=Jianbo Ye;126456]\r\n\r\nThe average helps, but not that much. LOL\r\n\r\n[/quote]\r\n\r\nwould you be so kind as to point us in the right direction? I would understand if you don't want to share your secret sauce\r\n\r\n[/quote]\r\n\r\nI am also wondering what other leaderboard runners are doing now! This is actually my first kaggle, and I learnt from them by researching how they win other competitions. \r\n\r\n\r\n[/quote]\r\n\r\nWhy did you join 5 years ago but not participate until now :p",
    "126494": "[quote=SteveKane;126493]\r\n\r\n[quote=Jianbo Ye;126491]\r\n\r\n[quote=SteveKane;126485]\r\n\r\n[quote=Jianbo Ye;126456]\r\n\r\nThe average helps, but not that much. LOL\r\n\r\n[/quote]\r\n\r\nwould you be so kind as to point us in the right direction? I would understand if you don't want to share your secret sauce\r\n\r\n[/quote]\r\n\r\nI am also wondering what other leaderboard runners are doing now! This is actually my first kaggle, and I learnt from them by researching how they win other competitions. \r\n\r\n\r\n[/quote]\r\n\r\nWhy did you join 5 years ago but not participate until now :p\r\n\r\n\r\n[/quote]\r\n\r\nGood question, I don't know either. But now I found Kaggle is too addictive, I can't help but continuously work on the problem."
  },
  "source": "meta"
}