{
  "id": 21365,
  "title": "Interactive exploratory analysis using a Shiny app",
  "url": "/competitions/facebook-v-predicting-check-ins/discussion/21365",
  "author_name": "",
  "post_date": "2016-06-01T15:56:53.540Z",
  "votes": 24,
  "comment_count": 12,
  "views": 2942,
  "content": "<p>Hi all,</p>\n\n<p>I have deployed a basic exploratory analysis app on shinyapps.io. The source code is posted as well in case my monthly quota runs out. The app covers about 5% (4783 unique place ids) of the place ids and 8% of the train data since there is a 100MB upload limit on shinyapps.io. The subset was selected by considering all place ids that belong to 5,000 random observations. Consequently, places with more observations are more likely to be contained in the subset. To run this locally you would need to create and convert the data file yourself (trainSubset.rds).</p>\n\n<p>I will be happy to consider any suggested improvements.</p>\n\n<p>Link: <a href=\"https://tvdwiele.shinyapps.io/Facebook-V/\">https://tvdwiele.shinyapps.io/Facebook-V/</a></p>\n\n<p>Enjoy :)</p>\n\n<p>Tom</p>",
  "messages": [
    {
      "id": "122150",
      "postDate": "06/01/2016 15:56:53",
      "content": "<p>Hi all,</p>\n\n<p>I have deployed a basic exploratory analysis app on shinyapps.io. The source code is posted as well in case my monthly quota runs out. The app covers about 5% (4783 unique place ids) of the place ids and 8% of the train data since there is a 100MB upload limit on shinyapps.io. The subset was selected by considering all place ids that belong to 5,000 random observations. Consequently, places with more observations are more likely to be contained in the subset. To run this locally you would need to create and convert the data file yourself (trainSubset.rds).</p>\n\n<p>I will be happy to consider any suggested improvements.</p>\n\n<p>Link: <a href=\"https://tvdwiele.shinyapps.io/Facebook-V/\">https://tvdwiele.shinyapps.io/Facebook-V/</a></p>\n\n<p>Enjoy :)</p>\n\n<p>Tom</p>",
      "rawMarkdown": "Hi all,\r\n\r\nI have deployed a basic exploratory analysis app on shinyapps.io. The source code is posted as well in case my monthly quota runs out. The app covers about 5% (4783 unique place ids) of the place ids and 8% of the train data since there is a 100MB upload limit on shinyapps.io. The subset was selected by considering all place ids that belong to 5,000 random observations. Consequently, places with more observations are more likely to be contained in the subset. To run this locally you would need to create and convert the data file yourself (trainSubset.rds).\r\n\r\nI will be happy to consider any suggested improvements.\r\n\r\nLink: https://tvdwiele.shinyapps.io/Facebook-V/\r\n\r\nEnjoy :)\r\n\r\nTom",
      "votes": null
    },
    {
      "id": "122157",
      "postDate": "06/01/2016 16:38:16",
      "content": "<p>Nice, app --- easy to quickly explore a few things. I liked it for the density plot of Hour :)</p>",
      "rawMarkdown": "Nice, app --- easy to quickly explore a few things. I liked it for the density plot of Hour :)",
      "votes": null
    },
    {
      "id": "122158",
      "postDate": "06/01/2016 16:46:52",
      "content": "<p>[quote=Tom Van de Wiele;122150]</p>\n\n<p>Hi all,</p>\n\n<p>I have deployed a basic exploratory analysis app on shinyapps.io. The source code is posted as well in case my monthly quota runs out. The app covers about 5% (4783 unique place ids) of the place ids and 8% of the train data since there is a 100MB upload limit on shinyapps.io. The subset was selected by considering all place ids that belong to 5,000 random observations. Consequently, places with more observations are more likely to be contained in the subset. To run this locally you would need to create and convert the data file yourself (trainSubset.rds).</p>\n\n<p>I will be happy to consider any suggested improvements.</p>\n\n<p>Link: <a href=\"https://tvdwiele.shinyapps.io/Facebook-V/\">https://tvdwiele.shinyapps.io/Facebook-V/</a></p>\n\n<p>Enjoy :)</p>\n\n<p>Tom</p>\n\n<p>[/quote]</p>\n\n<p>Nice work, an obvious extension is to show two two place_id to see the separability.</p>",
      "rawMarkdown": "[quote=Tom Van de Wiele;122150]\r\n\r\nHi all,\r\n\r\nI have deployed a basic exploratory analysis app on shinyapps.io. The source code is posted as well in case my monthly quota runs out. The app covers about 5% (4783 unique place ids) of the place ids and 8% of the train data since there is a 100MB upload limit on shinyapps.io. The subset was selected by considering all place ids that belong to 5,000 random observations. Consequently, places with more observations are more likely to be contained in the subset. To run this locally you would need to create and convert the data file yourself (trainSubset.rds).\r\n\r\nI will be happy to consider any suggested improvements.\r\n\r\nLink: https://tvdwiele.shinyapps.io/Facebook-V/\r\n\r\nEnjoy :)\r\n\r\nTom\r\n\r\n[/quote]\r\n\r\nNice work, an obvious extension is to show two two place_id to see the separability.",
      "votes": null
    },
    {
      "id": "122161",
      "postDate": "06/01/2016 16:57:26",
      "content": "<p>[quote=w_________g;122158]</p>\n\n<p>Nice work, an obvious extension is to show two two place_id to see the separability.</p>\n\n<p>[/quote]</p>\n\n<p>Thanks! As for your suggestion: I'm not sure that would be the best extension since there are so many possible place ids and I can only upload a fraction of the data. The separability is probably best analyzed in a a high dimensional space and that's not what this app was built for. </p>",
      "rawMarkdown": "[quote=w_________g;122158]\r\n\r\nNice work, an obvious extension is to show two two place_id to see the separability.\r\n\r\n[/quote]\r\n\r\nThanks! As for your suggestion: I'm not sure that would be the best extension since there are so many possible place ids and I can only upload a fraction of the data. The separability is probably best analyzed in a a high dimensional space and that's not what this app was built for.",
      "votes": null
    },
    {
      "id": "122163",
      "postDate": "06/01/2016 17:08:28",
      "content": "<p>[quote=Tom Van de Wiele;122161]</p>\n\n<p>[quote=w_________g;122158]</p>\n\n<p>Nice work, an obvious extension is to show two two place_id to see the separability.</p>\n\n<p>[/quote]</p>\n\n<p>Thanks! As for your suggestion: I'm not sure that would be the best extension since there are so many possible place ids and I can only upload a fraction of the data. The separability is probably best analyzed in a a high dimensional space and that's not what this app was built for. </p>\n\n<p>[/quote]\nThanks!</p>",
      "rawMarkdown": "[quote=Tom Van de Wiele;122161]\r\n\r\n[quote=w_________g;122158]\r\n\r\nNice work, an obvious extension is to show two two place_id to see the separability.\r\n\r\n[/quote]\r\n\r\nThanks! As for your suggestion: I'm not sure that would be the best extension since there are so many possible place ids and I can only upload a fraction of the data. The separability is probably best analyzed in a a high dimensional space and that's not what this app was built for. \r\n\r\n[/quote]\r\nThanks!",
      "votes": null
    },
    {
      "id": "122188",
      "postDate": "06/01/2016 23:24:33",
      "content": "<p>Great work. Might be cool to be able to see plots like hour vs accuracy for each place id.</p>",
      "rawMarkdown": "Great work. Might be cool to be able to see plots like hour vs accuracy for each place id.",
      "votes": null
    },
    {
      "id": "122210",
      "postDate": "06/02/2016 03:40:59",
      "content": "<p>Excellent suggestion Jared, thanks!</p>\n\n<p>The changes have been uploaded and the source files above were modified as well. I also added the option to look at the log transformed accuracy and fixed some minor bugs.</p>",
      "rawMarkdown": "Excellent suggestion Jared, thanks!\r\n\r\nThe changes have been uploaded and the source files above were modified as well. I also added the option to look at the log transformed accuracy and fixed some minor bugs.",
      "votes": null
    },
    {
      "id": "122211",
      "postDate": "06/02/2016 03:49:16",
      "content": "<p>And Wow, Jared, it seems like your suggestion may contain a signal.</p>\n\n<p>Edit: I added a violin plot comparison option and it seems like I was too eager to find a pattern between hour of the day and the accuracy.</p>",
      "rawMarkdown": "And Wow, Jared, it seems like your suggestion may contain a signal.\r\n\r\nEdit: I added a violin plot comparison option and it seems like I was too eager to find a pattern between hour of the day and the accuracy.",
      "votes": null
    },
    {
      "id": "122265",
      "postDate": "06/02/2016 15:25:15",
      "content": "<p>Hi Tom, Brilliant work \nIn the plot data what is the &quot;timeHalf&quot; First time part  &amp; Last time part</p>",
      "rawMarkdown": "Hi Tom, Brilliant work \r\nIn the plot data what is the \"timeHalf\" First time part  & Last time part",
      "votes": null
    },
    {
      "id": "122269",
      "postDate": "06/02/2016 15:48:18",
      "content": "<p>Radhakrishnan, that split allows the user to investigate time effects (particularly interesting for the accuracy). Based on the slider you can split the data at a variable time cutoff and repeat the analyses. It was a small addition I made this morning. The change is also reflected in the source files.</p>",
      "rawMarkdown": "Radhakrishnan, that split allows the user to investigate time effects (particularly interesting for the accuracy). Based on the slider you can split the data at a variable time cutoff and repeat the analyses. It was a small addition I made this morning. The change is also reflected in the source files.",
      "votes": null
    },
    {
      "id": "124838",
      "postDate": "06/22/2016 17:47:35",
      "content": "<p>[quote=Tom Van de Wiele;122211]</p>\n\n<p>And Wow, Jared, it seems like your suggestion may contain a signal.</p>\n\n<p>Edit: I added a violin plot comparison option and it seems like I was too eager to find a pattern between hour of the day and the accuracy.</p>\n\n<p>[/quote]</p>\n\n<p>I don't see the signal, can you elaborate?</p>\n\n<p>Or, can you give me a place id where it is obvious?</p>",
      "rawMarkdown": "[quote=Tom Van de Wiele;122211]\r\n\r\nAnd Wow, Jared, it seems like your suggestion may contain a signal.\r\n\r\nEdit: I added a violin plot comparison option and it seems like I was too eager to find a pattern between hour of the day and the accuracy.\r\n\r\n[/quote]\r\n\r\n\r\nI don't see the signal, can you elaborate?\r\n\r\nOr, can you give me a place id where it is obvious?",
      "votes": null
    },
    {
      "id": "124841",
      "postDate": "06/22/2016 17:59:16",
      "content": "<p>I was wrong, there is no obvious pattern between the hour of the day and the accuracy.</p>",
      "rawMarkdown": "I was wrong, there is no obvious pattern between the hour of the day and the accuracy.",
      "votes": null
    },
    {
      "id": "124922",
      "postDate": "06/23/2016 13:50:53",
      "content": "<p>Brilliant!!!</p>",
      "rawMarkdown": "Brilliant!!!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 122157,
      "author_name": "zeraes",
      "author_url": "",
      "post_date": "06/01/2016 16:38:16",
      "content": "<p>Nice, app --- easy to quickly explore a few things. I liked it for the density plot of Hour :)</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 122158,
      "author_name": "weizhezhang",
      "author_url": "",
      "post_date": "06/01/2016 16:46:52",
      "content": "<p>[quote=Tom Van de Wiele;122150]</p>\n\n<p>Hi all,</p>\n\n<p>I have deployed a basic exploratory analysis app on shinyapps.io. The source code is posted as well in case my monthly quota runs out. The app covers about 5% (4783 unique place ids) of the place ids and 8% of the train data since there is a 100MB upload limit on shinyapps.io. The subset was selected by considering all place ids that belong to 5,000 random observations. Consequently, places with more observations are more likely to be contained in the subset. To run this locally you would need to create and convert the data file yourself (trainSubset.rds).</p>\n\n<p>I will be happy to consider any suggested improvements.</p>\n\n<p>Link: <a href=\"https://tvdwiele.shinyapps.io/Facebook-V/\">https://tvdwiele.shinyapps.io/Facebook-V/</a></p>\n\n<p>Enjoy :)</p>\n\n<p>Tom</p>\n\n<p>[/quote]</p>\n\n<p>Nice work, an obvious extension is to show two two place_id to see the separability.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 122161,
      "author_name": "tvdwiele",
      "author_url": "",
      "post_date": "06/01/2016 16:57:26",
      "content": "<p>[quote=w_________g;122158]</p>\n\n<p>Nice work, an obvious extension is to show two two place_id to see the separability.</p>\n\n<p>[/quote]</p>\n\n<p>Thanks! As for your suggestion: I'm not sure that would be the best extension since there are so many possible place ids and I can only upload a fraction of the data. The separability is probably best analyzed in a a high dimensional space and that's not what this app was built for. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 122163,
      "author_name": "weizhezhang",
      "author_url": "",
      "post_date": "06/01/2016 17:08:28",
      "content": "<p>[quote=Tom Van de Wiele;122161]</p>\n\n<p>[quote=w_________g;122158]</p>\n\n<p>Nice work, an obvious extension is to show two two place_id to see the separability.</p>\n\n<p>[/quote]</p>\n\n<p>Thanks! As for your suggestion: I'm not sure that would be the best extension since there are so many possible place ids and I can only upload a fraction of the data. The separability is probably best analyzed in a a high dimensional space and that's not what this app was built for. </p>\n\n<p>[/quote]\nThanks!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 122188,
      "author_name": "jturkewitz",
      "author_url": "",
      "post_date": "06/01/2016 23:24:33",
      "content": "<p>Great work. Might be cool to be able to see plots like hour vs accuracy for each place id.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 122210,
      "author_name": "tvdwiele",
      "author_url": "",
      "post_date": "06/02/2016 03:40:59",
      "content": "<p>Excellent suggestion Jared, thanks!</p>\n\n<p>The changes have been uploaded and the source files above were modified as well. I also added the option to look at the log transformed accuracy and fixed some minor bugs.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 122211,
      "author_name": "tvdwiele",
      "author_url": "",
      "post_date": "06/02/2016 03:49:16",
      "content": "<p>And Wow, Jared, it seems like your suggestion may contain a signal.</p>\n\n<p>Edit: I added a violin plot comparison option and it seems like I was too eager to find a pattern between hour of the day and the accuracy.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 122265,
      "author_name": "rmsgrk",
      "author_url": "",
      "post_date": "06/02/2016 15:25:15",
      "content": "<p>Hi Tom, Brilliant work \nIn the plot data what is the &quot;timeHalf&quot; First time part  &amp; Last time part</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 122269,
      "author_name": "tvdwiele",
      "author_url": "",
      "post_date": "06/02/2016 15:48:18",
      "content": "<p>Radhakrishnan, that split allows the user to investigate time effects (particularly interesting for the accuracy). Based on the slider you can split the data at a variable time cutoff and repeat the analyses. It was a small addition I made this morning. The change is also reflected in the source files.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 124838,
      "author_name": "radustoicescu",
      "author_url": "",
      "post_date": "06/22/2016 17:47:35",
      "content": "<p>[quote=Tom Van de Wiele;122211]</p>\n\n<p>And Wow, Jared, it seems like your suggestion may contain a signal.</p>\n\n<p>Edit: I added a violin plot comparison option and it seems like I was too eager to find a pattern between hour of the day and the accuracy.</p>\n\n<p>[/quote]</p>\n\n<p>I don't see the signal, can you elaborate?</p>\n\n<p>Or, can you give me a place id where it is obvious?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 124841,
      "author_name": "tvdwiele",
      "author_url": "",
      "post_date": "06/22/2016 17:59:16",
      "content": "<p>I was wrong, there is no obvious pattern between the hour of the day and the accuracy.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 124922,
      "author_name": "pietromarinelli",
      "author_url": "",
      "post_date": "06/23/2016 13:50:53",
      "content": "<p>Brilliant!!!</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "122150": "Hi all,\r\n\r\nI have deployed a basic exploratory analysis app on shinyapps.io. The source code is posted as well in case my monthly quota runs out. The app covers about 5% (4783 unique place ids) of the place ids and 8% of the train data since there is a 100MB upload limit on shinyapps.io. The subset was selected by considering all place ids that belong to 5,000 random observations. Consequently, places with more observations are more likely to be contained in the subset. To run this locally you would need to create and convert the data file yourself (trainSubset.rds).\r\n\r\nI will be happy to consider any suggested improvements.\r\n\r\nLink: https://tvdwiele.shinyapps.io/Facebook-V/\r\n\r\nEnjoy :)\r\n\r\nTom",
    "122157": "Nice, app --- easy to quickly explore a few things. I liked it for the density plot of Hour :)",
    "122158": "[quote=Tom Van de Wiele;122150]\r\n\r\nHi all,\r\n\r\nI have deployed a basic exploratory analysis app on shinyapps.io. The source code is posted as well in case my monthly quota runs out. The app covers about 5% (4783 unique place ids) of the place ids and 8% of the train data since there is a 100MB upload limit on shinyapps.io. The subset was selected by considering all place ids that belong to 5,000 random observations. Consequently, places with more observations are more likely to be contained in the subset. To run this locally you would need to create and convert the data file yourself (trainSubset.rds).\r\n\r\nI will be happy to consider any suggested improvements.\r\n\r\nLink: https://tvdwiele.shinyapps.io/Facebook-V/\r\n\r\nEnjoy :)\r\n\r\nTom\r\n\r\n[/quote]\r\n\r\nNice work, an obvious extension is to show two two place_id to see the separability.",
    "122161": "[quote=w_________g;122158]\r\n\r\nNice work, an obvious extension is to show two two place_id to see the separability.\r\n\r\n[/quote]\r\n\r\nThanks! As for your suggestion: I'm not sure that would be the best extension since there are so many possible place ids and I can only upload a fraction of the data. The separability is probably best analyzed in a a high dimensional space and that's not what this app was built for.",
    "122163": "[quote=Tom Van de Wiele;122161]\r\n\r\n[quote=w_________g;122158]\r\n\r\nNice work, an obvious extension is to show two two place_id to see the separability.\r\n\r\n[/quote]\r\n\r\nThanks! As for your suggestion: I'm not sure that would be the best extension since there are so many possible place ids and I can only upload a fraction of the data. The separability is probably best analyzed in a a high dimensional space and that's not what this app was built for. \r\n\r\n[/quote]\r\nThanks!",
    "122188": "Great work. Might be cool to be able to see plots like hour vs accuracy for each place id.",
    "122210": "Excellent suggestion Jared, thanks!\r\n\r\nThe changes have been uploaded and the source files above were modified as well. I also added the option to look at the log transformed accuracy and fixed some minor bugs.",
    "122211": "And Wow, Jared, it seems like your suggestion may contain a signal.\r\n\r\nEdit: I added a violin plot comparison option and it seems like I was too eager to find a pattern between hour of the day and the accuracy.",
    "122265": "Hi Tom, Brilliant work \r\nIn the plot data what is the \"timeHalf\" First time part  & Last time part",
    "122269": "Radhakrishnan, that split allows the user to investigate time effects (particularly interesting for the accuracy). Based on the slider you can split the data at a variable time cutoff and repeat the analyses. It was a small addition I made this morning. The change is also reflected in the source files.",
    "124838": "[quote=Tom Van de Wiele;122211]\r\n\r\nAnd Wow, Jared, it seems like your suggestion may contain a signal.\r\n\r\nEdit: I added a violin plot comparison option and it seems like I was too eager to find a pattern between hour of the day and the accuracy.\r\n\r\n[/quote]\r\n\r\n\r\nI don't see the signal, can you elaborate?\r\n\r\nOr, can you give me a place id where it is obvious?",
    "124841": "I was wrong, there is no obvious pattern between the hour of the day and the accuracy.",
    "124922": "Brilliant!!!"
  },
  "source": "meta"
}