I am curious about the usage of Time.\nAre these locations that change over time? Perhaps a mobile food truck?
\n\nThen in relation to Accuracy\nHow is accuracy determined in the real world? Would it be in relationship to the distance the user is from a cell phone tower? Will more towers improve accuracy or diminish it?
", "messages": [ { "id": "119620", "postDate": "05/11/2016 22:31:27", "content": "I am curious about the usage of Time.\nAre these locations that change over time? Perhaps a mobile food truck?
\n\nThen in relation to Accuracy\nHow is accuracy determined in the real world? Would it be in relationship to the distance the user is from a cell phone tower? Will more towers improve accuracy or diminish it?
", "rawMarkdown": "I am curious about the usage of Time.\r\nAre these locations that change over time? Perhaps a mobile food truck?\r\n\r\nThen in relation to Accuracy\r\nHow is accuracy determined in the real world? Would it be in relationship to the distance the user is from a cell phone tower? Will more towers improve accuracy or diminish it?", "votes": null }, { "id": "119626", "postDate": "05/12/2016 00:01:26", "content": "From here:
\n\n\n", "rawMarkdown": "From [here][1]:\r\n\r\n> Note: Some of the columns, such as time and accuracy, are\r\n> intentionally left vague in their definitions. Please consider them as\r\n> part of the challenge.\r\n\r\n\r\n [1]: https://www.kaggle.com/c/facebook-v-predicting-check-ins/data", "votes": null }, { "id": "119653", "postDate": "05/12/2016 03:37:18", "content": "Note: Some of the columns, such as time and accuracy, are\n intentionally left vague in their definitions. Please consider them as\n part of the challenge.
\n
I guess the accuracy is the number of meters the GPS is accurate to. Depending on how you use your mobile, I found the below: \n3G iPhone w/ A-GPS ~ 8 meters\n3G iPhone w/ wifi ~ 74 meters\n3G iPhone w/ Cellular positioning ~ 600 meters
", "rawMarkdown": "I guess the accuracy is the number of meters the GPS is accurate to. Depending on how you use your mobile, I found the below: \r\n3G iPhone w/ A-GPS ~ 8 meters\r\n3G iPhone w/ wifi ~ 74 meters\r\n3G iPhone w/ Cellular positioning ~ 600 meters", "votes": null }, { "id": "120485", "postDate": "05/18/2016 15:53:17", "content": "There are several ways that you can measure user's location.
\n\n1) Network Based Method \nSometimes called "cell tower triangular" method. In short, your relative location to three cell towers in your service provider is approximated using the algorithm, then service provider can identify your location.
\n\n2) Handset(mobile phone) Based Method \nThis method uses GPS in cell phone, but not all handset has GPS.
\n\nThe location accuracy of method 1) and 2) and its combined are regulated by FCC due to E911 services.
\n\n3) Wi-Fi AP (Access Point) Based Method \nThe third method is when you connect to the Wi-Fi, the (cable or broadband) service provider has the knowledge of the location of AP, thus knows the location (maybe within x meters or so). But I often sit in Barnes and Nobles but connect to the next door bakery Wi-Fi since BN Wi-Fi is always busy. So is this accurate information of my location? :)
\n\n4) There are several other new ways to estimate the location. \nIf you use a certain VoIP call (if it actually uses a service category called IMS), there is another way to capture location in location service server. There are other recent proposals such as using beacon which uses bluetooth.
\n\n5) To know exactly where the user is ... \nThis is still a research topic. Once GPS penetration is higher, hybrid methods of handset based and Network Based will likely increase the accuracy for service providers.
\n\nBut for mobile apps, there are some standardized protocols exist. Mobile Apps still can use all the information that service provider can obtain through either Google or Apple or any another company (OS) interface, so called API.
\n\nToday is my first day in this competition and haven't read all the posts. Reading only some of forums, I just want to say, it is very unlikely even the service provider or facebook knows exact true location unless the data is gathered in some test setting. :) One way to know the true location with current technology is that if there is a payment information at a retail store. It will reveal user's true location most likely. ^^
", "rawMarkdown": "There are several ways that you can measure user's location. \r\n\r\n1) Network Based Method \r\nSometimes called \"cell tower triangular\" method. In short, your relative location to three cell towers in your service provider is approximated using the algorithm, then service provider can identify your location. \r\n\r\n2) Handset(mobile phone) Based Method \r\nThis method uses GPS in cell phone, but not all handset has GPS. \r\n\r\nThe location accuracy of method 1) and 2) and its combined are regulated by FCC due to E911 services. \r\n \r\n3) Wi-Fi AP (Access Point) Based Method \r\nThe third method is when you connect to the Wi-Fi, the (cable or broadband) service provider has the knowledge of the location of AP, thus knows the location (maybe within x meters or so). But I often sit in Barnes and Nobles but connect to the next door bakery Wi-Fi since BN Wi-Fi is always busy. So is this accurate information of my location? :) \r\n \r\n4) There are several other new ways to estimate the location. \r\nIf you use a certain VoIP call (if it actually uses a service category called IMS), there is another way to capture location in location service server. There are other recent proposals such as using beacon which uses bluetooth. \r\n \r\n5) To know exactly where the user is ... \r\nThis is still a research topic. Once GPS penetration is higher, hybrid methods of handset based and Network Based will likely increase the accuracy for service providers. \r\n \r\nBut for mobile apps, there are some standardized protocols exist. Mobile Apps still can use all the information that service provider can obtain through either Google or Apple or any another company (OS) interface, so called API. \r\n \r\nToday is my first day in this competition and haven't read all the posts. Reading only some of forums, I just want to say, it is very unlikely even the service provider or facebook knows exact true location unless the data is gathered in some test setting. :) One way to know the true location with current technology is that if there is a payment information at a retail store. It will reveal user's true location most likely. ^^", "votes": null }, { "id": "120486", "postDate": "05/18/2016 15:58:12", "content": "@Young I like the list.
\n\nIf the "errors" are systematic, true locations might be recovered. If they are random, might be harder. Thinking about those mechanisms you point out will be good food for thought while trying to determine what can be recovered. Thanks.
", "rawMarkdown": "Young I like the list.\r\n\r\nIf the \"errors\" are systematic, true locations might be recovered. If they are random, might be harder. Thinking about those mechanisms you point out will be good food for thought while trying to determine what can be recovered. Thanks.", "votes": null }, { "id": "120496", "postDate": "05/18/2016 16:50:14", "content": "@RyanBatt,
\n\nThere are few root causes of errors that I can think of now: \ni) you can search the cause of GPS error range. Probably, a simple example can be indoor accuracy vs. outdoor accuracy. \nii) how good a GPS chipset in your mobile and the equipment in service provider detecting it. (The performance(accuracy) of chipset and GPS equipment actually varies.)
\niii) if it is a network based method, line-of-sight is a subject of error. Line-of-sight means, if you are behind a mountain or tall building which blocks the signal from cell towers, the 2D location information accuracy may drop. \niv) As for the AP-base method, as I mentioned above, if a user is in a city or in a mall or any other location where many free Wi-Fi can be found at the same time in one location, then it will impact the accuracy depending on what accuracy means - distance or actual location as I mentioned above.
In short, the source of measurement errors are very broad from the law of physics, a matter of money (chipset/equipment quality) ,the density of Wi-Fi distribution & Wi-Fi user density, and whether user turns on location services.
\n\nA question that occurs to me is that should we assume multiple sources of errors and treat "accuracy" as a qualitative value, or "accuracy" still is the (empirical) quantitative measurement error in some method used by FB/service provider that is not known?
\n\nCheers,\nYoung
", "rawMarkdown": "RyanBatt, \r\n\r\nThere are few root causes of errors that I can think of now: \r\ni) you can search the cause of GPS error range. Probably, a simple example can be indoor accuracy vs. outdoor accuracy. \r\nii) how good a GPS chipset in your mobile and the equipment in service provider detecting it. (The performance(accuracy) of chipset and GPS equipment actually varies.) \r\niii) if it is a network based method, line-of-sight is a subject of error. Line-of-sight means, if you are behind a mountain or tall building which blocks the signal from cell towers, the 2D location information accuracy may drop. \r\niv) As for the AP-base method, as I mentioned above, if a user is in a city or in a mall or any other location where many free Wi-Fi can be found at the same time in one location, then it will impact the accuracy depending on what accuracy means - distance or actual location as I mentioned above. \r\n\r\nIn short, the source of measurement errors are very broad from the law of physics, a matter of money (chipset/equipment quality) ,the density of Wi-Fi distribution & Wi-Fi user density, and whether user turns on location services. \r\n \r\nA question that occurs to me is that should we assume multiple sources of errors and treat \"accuracy\" as a qualitative value, or \"accuracy\" still is the (empirical) quantitative measurement error in some method used by FB/service provider that is not known? \r\n \r\nCheers,\r\nYoung", "votes": null } ], "comments": [ { "id": 119626, "author_name": "carloshuertas", "author_url": "", "post_date": "05/12/2016 00:01:26", "content": "From here:
\n\n\n", "votes": null, "replies": [] }, { "id": 119653, "author_name": "kumarmankala", "author_url": "", "post_date": "05/12/2016 03:37:18", "content": "Note: Some of the columns, such as time and accuracy, are\n intentionally left vague in their definitions. Please consider them as\n part of the challenge.
\n
I guess the accuracy is the number of meters the GPS is accurate to. Depending on how you use your mobile, I found the below: \n3G iPhone w/ A-GPS ~ 8 meters\n3G iPhone w/ wifi ~ 74 meters\n3G iPhone w/ Cellular positioning ~ 600 meters
", "votes": null, "replies": [] }, { "id": 120485, "author_name": "ykim007", "author_url": "", "post_date": "05/18/2016 15:53:17", "content": "There are several ways that you can measure user's location.
\n\n1) Network Based Method \nSometimes called "cell tower triangular" method. In short, your relative location to three cell towers in your service provider is approximated using the algorithm, then service provider can identify your location.
\n\n2) Handset(mobile phone) Based Method \nThis method uses GPS in cell phone, but not all handset has GPS.
\n\nThe location accuracy of method 1) and 2) and its combined are regulated by FCC due to E911 services.
\n\n3) Wi-Fi AP (Access Point) Based Method \nThe third method is when you connect to the Wi-Fi, the (cable or broadband) service provider has the knowledge of the location of AP, thus knows the location (maybe within x meters or so). But I often sit in Barnes and Nobles but connect to the next door bakery Wi-Fi since BN Wi-Fi is always busy. So is this accurate information of my location? :)
\n\n4) There are several other new ways to estimate the location. \nIf you use a certain VoIP call (if it actually uses a service category called IMS), there is another way to capture location in location service server. There are other recent proposals such as using beacon which uses bluetooth.
\n\n5) To know exactly where the user is ... \nThis is still a research topic. Once GPS penetration is higher, hybrid methods of handset based and Network Based will likely increase the accuracy for service providers.
\n\nBut for mobile apps, there are some standardized protocols exist. Mobile Apps still can use all the information that service provider can obtain through either Google or Apple or any another company (OS) interface, so called API.
\n\nToday is my first day in this competition and haven't read all the posts. Reading only some of forums, I just want to say, it is very unlikely even the service provider or facebook knows exact true location unless the data is gathered in some test setting. :) One way to know the true location with current technology is that if there is a payment information at a retail store. It will reveal user's true location most likely. ^^
", "votes": null, "replies": [] }, { "id": 120486, "author_name": "zeraes", "author_url": "", "post_date": "05/18/2016 15:58:12", "content": "@Young I like the list.
\n\nIf the "errors" are systematic, true locations might be recovered. If they are random, might be harder. Thinking about those mechanisms you point out will be good food for thought while trying to determine what can be recovered. Thanks.
", "votes": null, "replies": [] }, { "id": 120496, "author_name": "ykim007", "author_url": "", "post_date": "05/18/2016 16:50:14", "content": "@RyanBatt,
\n\nThere are few root causes of errors that I can think of now: \ni) you can search the cause of GPS error range. Probably, a simple example can be indoor accuracy vs. outdoor accuracy. \nii) how good a GPS chipset in your mobile and the equipment in service provider detecting it. (The performance(accuracy) of chipset and GPS equipment actually varies.)
\niii) if it is a network based method, line-of-sight is a subject of error. Line-of-sight means, if you are behind a mountain or tall building which blocks the signal from cell towers, the 2D location information accuracy may drop. \niv) As for the AP-base method, as I mentioned above, if a user is in a city or in a mall or any other location where many free Wi-Fi can be found at the same time in one location, then it will impact the accuracy depending on what accuracy means - distance or actual location as I mentioned above.
In short, the source of measurement errors are very broad from the law of physics, a matter of money (chipset/equipment quality) ,the density of Wi-Fi distribution & Wi-Fi user density, and whether user turns on location services.
\n\nA question that occurs to me is that should we assume multiple sources of errors and treat "accuracy" as a qualitative value, or "accuracy" still is the (empirical) quantitative measurement error in some method used by FB/service provider that is not known?
\n\nCheers,\nYoung
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "119620": "I am curious about the usage of Time.\r\nAre these locations that change over time? Perhaps a mobile food truck?\r\n\r\nThen in relation to Accuracy\r\nHow is accuracy determined in the real world? Would it be in relationship to the distance the user is from a cell phone tower? Will more towers improve accuracy or diminish it?", "119626": "From [here][1]:\r\n\r\n> Note: Some of the columns, such as time and accuracy, are\r\n> intentionally left vague in their definitions. Please consider them as\r\n> part of the challenge.\r\n\r\n\r\n [1]: https://www.kaggle.com/c/facebook-v-predicting-check-ins/data", "119653": "I guess the accuracy is the number of meters the GPS is accurate to. Depending on how you use your mobile, I found the below: \r\n3G iPhone w/ A-GPS ~ 8 meters\r\n3G iPhone w/ wifi ~ 74 meters\r\n3G iPhone w/ Cellular positioning ~ 600 meters", "120485": "There are several ways that you can measure user's location. \r\n\r\n1) Network Based Method \r\nSometimes called \"cell tower triangular\" method. In short, your relative location to three cell towers in your service provider is approximated using the algorithm, then service provider can identify your location. \r\n\r\n2) Handset(mobile phone) Based Method \r\nThis method uses GPS in cell phone, but not all handset has GPS. \r\n\r\nThe location accuracy of method 1) and 2) and its combined are regulated by FCC due to E911 services. \r\n \r\n3) Wi-Fi AP (Access Point) Based Method \r\nThe third method is when you connect to the Wi-Fi, the (cable or broadband) service provider has the knowledge of the location of AP, thus knows the location (maybe within x meters or so). But I often sit in Barnes and Nobles but connect to the next door bakery Wi-Fi since BN Wi-Fi is always busy. So is this accurate information of my location? :) \r\n \r\n4) There are several other new ways to estimate the location. \r\nIf you use a certain VoIP call (if it actually uses a service category called IMS), there is another way to capture location in location service server. There are other recent proposals such as using beacon which uses bluetooth. \r\n \r\n5) To know exactly where the user is ... \r\nThis is still a research topic. Once GPS penetration is higher, hybrid methods of handset based and Network Based will likely increase the accuracy for service providers. \r\n \r\nBut for mobile apps, there are some standardized protocols exist. Mobile Apps still can use all the information that service provider can obtain through either Google or Apple or any another company (OS) interface, so called API. \r\n \r\nToday is my first day in this competition and haven't read all the posts. Reading only some of forums, I just want to say, it is very unlikely even the service provider or facebook knows exact true location unless the data is gathered in some test setting. :) One way to know the true location with current technology is that if there is a payment information at a retail store. It will reveal user's true location most likely. ^^", "120486": "Young I like the list.\r\n\r\nIf the \"errors\" are systematic, true locations might be recovered. If they are random, might be harder. Thinking about those mechanisms you point out will be good food for thought while trying to determine what can be recovered. Thanks.", "120496": "RyanBatt, \r\n\r\nThere are few root causes of errors that I can think of now: \r\ni) you can search the cause of GPS error range. Probably, a simple example can be indoor accuracy vs. outdoor accuracy. \r\nii) how good a GPS chipset in your mobile and the equipment in service provider detecting it. (The performance(accuracy) of chipset and GPS equipment actually varies.) \r\niii) if it is a network based method, line-of-sight is a subject of error. Line-of-sight means, if you are behind a mountain or tall building which blocks the signal from cell towers, the 2D location information accuracy may drop. \r\niv) As for the AP-base method, as I mentioned above, if a user is in a city or in a mall or any other location where many free Wi-Fi can be found at the same time in one location, then it will impact the accuracy depending on what accuracy means - distance or actual location as I mentioned above. \r\n\r\nIn short, the source of measurement errors are very broad from the law of physics, a matter of money (chipset/equipment quality) ,the density of Wi-Fi distribution & Wi-Fi user density, and whether user turns on location services. \r\n \r\nA question that occurs to me is that should we assume multiple sources of errors and treat \"accuracy\" as a qualitative value, or \"accuracy\" still is the (empirical) quantitative measurement error in some method used by FB/service provider that is not known? \r\n \r\nCheers,\r\nYoung" }, "source": "meta" }