{
  "id": 66290,
  "title": "Suggestion: Challenge in creating model",
  "url": "/competitions/ga-customer-revenue-prediction/discussion/66290",
  "author_name": "Gaurav Jha",
  "post_date": "2018-09-20T03:32:04.093000",
  "votes": 0,
  "comment_count": 12,
  "views": 0,
  "content": "<p>Hi All,</p>\n\n<p>I am facing a challenge while creating the model, as owing to the large number of categorical variables and their levels as well.</p>\n\n<p>I am using a HP Prodesk machine with Windows 10 pro (64-bit) and 8 GB RAM.</p>\n\n<p>Is there any other platforms where I can create and run my models, without any charge.</p>\n\n<p>Any suggestions welcome!!!!!</p>\n\n<p>Thanks in advance.</p>",
  "messages": [
    {
      "id": 390629,
      "postDate": "2018-09-20T14:45:40.897Z",
      "content": "<p>Kaggle kernel is definitly better for you. The only catch is that you have to ensure that your model runs under 6 hours (since I believe that is the time limit for the active session for a unique user). </p>\n\n<p>Other option could be if you remove windows from your system and move to Ubuntu. That way you would get most of your 8GB RAM (windows must be giving you around ~5GB Max for usage) and then train models on smaller samples of train data.</p>\n\n<p>Hope this helps.</p>",
      "rawMarkdown": "Kaggle kernel is definitly better for you. The only catch is that you have to ensure that your model runs under 6 hours (since I believe that is the time limit for the active session for a unique user). \n\nOther option could be if you remove windows from your system and move to Ubuntu. That way you would get most of your 8GB RAM (windows must be giving you around ~5GB Max for usage) and then train models on smaller samples of train data.\n\nHope this helps.",
      "replies": [
        {
          "id": 390912,
          "postDate": "2018-09-21T03:06:09.503Z",
          "content": "<p>Thanks for the reply. I was trying some dimension reduction on the dataset, but the kaggle kernel was failing as well. I will try to reduce the data some more and try again.</p>",
          "rawMarkdown": "Thanks for the reply. I was trying some dimension reduction on the dataset, but the kaggle kernel was failing as well. I will try to reduce the data some more and try again."
        }
      ]
    },
    {
      "id": 390360,
      "postDate": "2018-09-20T04:52:28.430Z",
      "content": "<p>You could use kaggle kernels, which provide 16 GB RAM...</p>",
      "rawMarkdown": "You could use kaggle kernels, which provide 16 GB RAM...",
      "replies": [
        {
          "id": 390916,
          "postDate": "2018-09-21T03:07:48.690Z",
          "content": "<p>Thanks. I tried this with full data and the kernel was failing. I will reduce the data and try again.</p>",
          "rawMarkdown": "Thanks. I tried this with full data and the kernel was failing. I will reduce the data and try again."
        }
      ]
    },
    {
      "id": 390346,
      "postDate": "2018-09-20T04:15:44.450Z",
      "content": "<p>For tree based model its better if you can use label encoding </p>",
      "rawMarkdown": "For tree based model its better if you can use label encoding ",
      "replies": [
        {
          "id": 390910,
          "postDate": "2018-09-21T03:03:38.197Z",
          "content": "<p>My bad, I forgot to mention, I am using a Regression model. Thanks anyway for the suggestion.</p>",
          "rawMarkdown": "My bad, I forgot to mention, I am using a Regression model. Thanks anyway for the suggestion."
        }
      ]
    },
    {
      "id": 390332,
      "postDate": "2018-09-20T03:32:04.093Z",
      "content": "<p>Hi All,</p>\n\n<p>I am facing a challenge while creating the model, as owing to the large number of categorical variables and their levels as well.</p>\n\n<p>I am using a HP Prodesk machine with Windows 10 pro (64-bit) and 8 GB RAM.</p>\n\n<p>Is there any other platforms where I can create and run my models, without any charge.</p>\n\n<p>Any suggestions welcome!!!!!</p>\n\n<p>Thanks in advance.</p>",
      "rawMarkdown": "Hi All,\n\nI am facing a challenge while creating the model, as owing to the large number of categorical variables and their levels as well.\n\nI am using a HP Prodesk machine with Windows 10 pro (64-bit) and 8 GB RAM.\n\nIs there any other platforms where I can create and run my models, without any charge.\n\nAny suggestions welcome!!!!!\n\nThanks in advance."
    },
    {
      "id": 390634,
      "postDate": "2018-09-20T15:00:53.433Z",
      "rawMarkdown": "",
      "isDeleted": true,
      "replies": [
        {
          "id": 390914,
          "postDate": "2018-09-21T03:06:55.997Z",
          "content": "<p>Thanks. I read thru this and one of the eligibility criteria is that a successful submission needs to be made. I am struggling with my first submission.</p>",
          "rawMarkdown": "Thanks. I read thru this and one of the eligibility criteria is that a successful submission needs to be made. I am struggling with my first submission."
        },
        {
          "id": 390987,
          "postDate": "2018-09-21T05:53:05.790Z",
          "content": "<p>You may make a submission by training your model on a very small sample...the small sample can be done while you read the data. That way once you have made the submission, you use the GC credits for building a bigger model.</p>",
          "rawMarkdown": "You may make a submission by training your model on a very small sample...the small sample can be done while you read the data. That way once you have made the submission, you use the GC credits for building a bigger model."
        }
      ]
    },
    {
      "id": 390584,
      "postDate": "2018-09-20T13:18:15.473Z",
      "rawMarkdown": "",
      "votes": 2,
      "isDeleted": true,
      "replies": [
        {
          "id": 390911,
          "postDate": "2018-09-21T03:04:27.940Z",
          "content": "<p>Thanks. I will try <a href=\"https://machinelabs.ai/\">https://machinelabs.ai/</a>.</p>",
          "rawMarkdown": "Thanks. I will try https://machinelabs.ai/."
        },
        {
          "id": 390917,
          "postDate": "2018-09-21T03:10:35.470Z",
          "content": "<p>It looks like <a href=\"https://machinelabs.ai/\">https://machinelabs.ai/</a> is only for python. I did not see any option to write R scripts in it. Can you please suggest if R is supported or is it just for Python.</p>",
          "rawMarkdown": "It looks like https://machinelabs.ai/ is only for python. I did not see any option to write R scripts in it. Can you please suggest if R is supported or is it just for Python."
        }
      ]
    }
  ],
  "comments": [
    {
      "id": 390629,
      "author_name": "sandy1112",
      "author_url": "",
      "post_date": "2018-09-20T14:45:40.897000",
      "content": "<p>Kaggle kernel is definitly better for you. The only catch is that you have to ensure that your model runs under 6 hours (since I believe that is the time limit for the active session for a unique user). </p>\n\n<p>Other option could be if you remove windows from your system and move to Ubuntu. That way you would get most of your 8GB RAM (windows must be giving you around ~5GB Max for usage) and then train models on smaller samples of train data.</p>\n\n<p>Hope this helps.</p>",
      "votes": 0,
      "replies": [
        {
          "id": 390912,
          "author_name": "Gaurav Jha",
          "author_url": "",
          "post_date": "2018-09-21T03:06:09.503000",
          "content": "<p>Thanks for the reply. I was trying some dimension reduction on the dataset, but the kaggle kernel was failing as well. I will try to reduce the data some more and try again.</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 390360,
      "author_name": "Andrey Lukyanenko",
      "author_url": "",
      "post_date": "2018-09-20T04:52:28.430000",
      "content": "<p>You could use kaggle kernels, which provide 16 GB RAM...</p>",
      "votes": 0,
      "replies": [
        {
          "id": 390916,
          "author_name": "Gaurav Jha",
          "author_url": "",
          "post_date": "2018-09-21T03:07:48.690000",
          "content": "<p>Thanks. I tried this with full data and the kernel was failing. I will reduce the data and try again.</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 390346,
      "author_name": "redshellspy",
      "author_url": "",
      "post_date": "2018-09-20T04:15:44.450000",
      "content": "<p>For tree based model its better if you can use label encoding </p>",
      "votes": 0,
      "replies": [
        {
          "id": 390910,
          "author_name": "Gaurav Jha",
          "author_url": "",
          "post_date": "2018-09-21T03:03:38.197000",
          "content": "<p>My bad, I forgot to mention, I am using a Regression model. Thanks anyway for the suggestion.</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 390634,
      "author_name": "",
      "author_url": "",
      "post_date": "2018-09-20T15:00:53.433000",
      "content": "",
      "votes": 0,
      "replies": [
        {
          "id": 390914,
          "author_name": "Gaurav Jha",
          "author_url": "",
          "post_date": "2018-09-21T03:06:55.997000",
          "content": "<p>Thanks. I read thru this and one of the eligibility criteria is that a successful submission needs to be made. I am struggling with my first submission.</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 390987,
          "author_name": "sandy1112",
          "author_url": "",
          "post_date": "2018-09-21T05:53:05.790000",
          "content": "<p>You may make a submission by training your model on a very small sample...the small sample can be done while you read the data. That way once you have made the submission, you use the GC credits for building a bigger model.</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 390584,
      "author_name": "",
      "author_url": "",
      "post_date": "2018-09-20T13:18:15.473000",
      "content": "",
      "votes": 2,
      "replies": [
        {
          "id": 390911,
          "author_name": "Gaurav Jha",
          "author_url": "",
          "post_date": "2018-09-21T03:04:27.940000",
          "content": "<p>Thanks. I will try <a href=\"https://machinelabs.ai/\">https://machinelabs.ai/</a>.</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 390917,
          "author_name": "Gaurav Jha",
          "author_url": "",
          "post_date": "2018-09-21T03:10:35.470000",
          "content": "<p>It looks like <a href=\"https://machinelabs.ai/\">https://machinelabs.ai/</a> is only for python. I did not see any option to write R scripts in it. Can you please suggest if R is supported or is it just for Python.</p>",
          "votes": 0,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "390629": "Kaggle kernel is definitly better for you. The only catch is that you have to ensure that your model runs under 6 hours (since I believe that is the time limit for the active session for a unique user). \n\nOther option could be if you remove windows from your system and move to Ubuntu. That way you would get most of your 8GB RAM (windows must be giving you around ~5GB Max for usage) and then train models on smaller samples of train data.\n\nHope this helps.",
    "390360": "You could use kaggle kernels, which provide 16 GB RAM...",
    "390346": "For tree based model its better if you can use label encoding ",
    "390332": "Hi All,\n\nI am facing a challenge while creating the model, as owing to the large number of categorical variables and their levels as well.\n\nI am using a HP Prodesk machine with Windows 10 pro (64-bit) and 8 GB RAM.\n\nIs there any other platforms where I can create and run my models, without any charge.\n\nAny suggestions welcome!!!!!\n\nThanks in advance.",
    "390634": "",
    "390584": ""
  }
}