Hi, in training set and testing set there both exists an "accuracy" column, in the data description page it states that "accuracy: location accuracy".
\n\nMy interpretation of this is that - accuracy is solely related to x,y coordinates (locations), for example a person may do facebook checkin at location (x1, y1), but in dataset it is recorded that this guy is at (x2, y2). And accuracy column is generated as a measure of difference between (x1, y1) and (x2, y2).
\n\nIs the above interpretation correct? If not - then why does the data description page states "accuracy: location accuracy"?
\n\nThanks in advance.
", "messages": [ { "id": "119902", "postDate": "05/13/2016 16:38:21", "content": "Hi, in training set and testing set there both exists an "accuracy" column, in the data description page it states that "accuracy: location accuracy".
\n\nMy interpretation of this is that - accuracy is solely related to x,y coordinates (locations), for example a person may do facebook checkin at location (x1, y1), but in dataset it is recorded that this guy is at (x2, y2). And accuracy column is generated as a measure of difference between (x1, y1) and (x2, y2).
\n\nIs the above interpretation correct? If not - then why does the data description page states "accuracy: location accuracy"?
\n\nThanks in advance.
", "rawMarkdown": "Hi, in training set and testing set there both exists an \"accuracy\" column, in the data description page it states that \"accuracy: location accuracy\".\r\n\r\nMy interpretation of this is that - accuracy is solely related to x,y coordinates (locations), for example a person may do facebook checkin at location (x1, y1), but in dataset it is recorded that this guy is at (x2, y2). And accuracy column is generated as a measure of difference between (x1, y1) and (x2, y2).\r\n\r\nIs the above interpretation correct? If not - then why does the data description page states \"accuracy: location accuracy\"? \r\n\r\nThanks in advance.", "votes": null }, { "id": "119908", "postDate": "05/13/2016 17:18:26", "content": "I'm interpreting it as a measure of the reliability of the (x,y) values. ie the higher the accuracy, the more confidence we can have that (x2,y2) on the data are close to actual (unknown) location (x1,y1) of the user. The data page notes that the definitions are left "intentionally vague" so I suspect it's up to us to figure out the implications of high/low accuracy.
\n\nSome of the data exploration done by others here seems to confirm the idea that high accuracy => better x,y values (as opposed to low accuracy values being better).
\n\nI haven't yet been able to do anything useful with accuracy (despite having tried a few ideas), so I'm be interested to see what people end up doing with it.
", "rawMarkdown": "I'm interpreting it as a measure of the reliability of the (x,y) values. ie the higher the accuracy, the more confidence we can have that (x2,y2) on the data are close to actual (unknown) location (x1,y1) of the user. The data page notes that the definitions are left \"intentionally vague\" so I suspect it's up to us to figure out the implications of high/low accuracy.\r\n\r\nSome of the data exploration done by others here seems to confirm the idea that high accuracy => better x,y values (as opposed to low accuracy values being better).\r\n\r\nI haven't yet been able to do anything useful with accuracy (despite having tried a few ideas), so I'm be interested to see what people end up doing with it.", "votes": null }, { "id": "119909", "postDate": "05/13/2016 17:24:52", "content": "Thanks Peatle - your comment confirms my thoughts.
\n\nI'm not 100% sure how accuracy is calculated, but what I'm sure now is that accuracy is ONLY related to x,y coordinates - it has nothing to do with time and place_id.
\n\nIf there are different opinion welcome to chime in. Thanks again.
", "rawMarkdown": "Thanks Peatle - your comment confirms my thoughts.\r\n\r\nI'm not 100% sure how accuracy is calculated, but what I'm sure now is that accuracy is ONLY related to x,y coordinates - it has nothing to do with time and place_id.\r\n\r\nIf there are different opinion welcome to chime in. Thanks again.", "votes": null }, { "id": "124374", "postDate": "06/17/2016 16:48:09", "content": "I agree with the accuracy discussions above.
\n\nWe do have three peaks in the accuracy histogram. One way to understand them is that accuracy of (x,y) is dependent on what technology is used in the phone for positioning. The peaks could correspond to usage of GPS, WiFi, and cellular position.
\n\nThis paper gives a good description (e.g. see Figure 3): http://www.paulzandbergen.com/PUBLICATIONS_files/Zandbergen_TGIS_2009.pdf
\n\nThe question is how to use this knowledge...
", "rawMarkdown": "I agree with the accuracy discussions above. \r\n\r\nWe do have three peaks in the accuracy histogram. One way to understand them is that accuracy of (x,y) is dependent on what technology is used in the phone for positioning. The peaks could correspond to usage of GPS, WiFi, and cellular position. \r\n\r\nThis paper gives a good description (e.g. see Figure 3): http://www.paulzandbergen.com/PUBLICATIONS_files/Zandbergen_TGIS_2009.pdf\r\n\r\nThe question is how to use this knowledge...", "votes": null } ], "comments": [ { "id": 119908, "author_name": "peatle", "author_url": "", "post_date": "05/13/2016 17:18:26", "content": "I'm interpreting it as a measure of the reliability of the (x,y) values. ie the higher the accuracy, the more confidence we can have that (x2,y2) on the data are close to actual (unknown) location (x1,y1) of the user. The data page notes that the definitions are left "intentionally vague" so I suspect it's up to us to figure out the implications of high/low accuracy.
\n\nSome of the data exploration done by others here seems to confirm the idea that high accuracy => better x,y values (as opposed to low accuracy values being better).
\n\nI haven't yet been able to do anything useful with accuracy (despite having tried a few ideas), so I'm be interested to see what people end up doing with it.
", "votes": null, "replies": [] }, { "id": 119909, "author_name": "howardxie", "author_url": "", "post_date": "05/13/2016 17:24:52", "content": "Thanks Peatle - your comment confirms my thoughts.
\n\nI'm not 100% sure how accuracy is calculated, but what I'm sure now is that accuracy is ONLY related to x,y coordinates - it has nothing to do with time and place_id.
\n\nIf there are different opinion welcome to chime in. Thanks again.
", "votes": null, "replies": [] }, { "id": 124374, "author_name": "andreasjohnsson", "author_url": "", "post_date": "06/17/2016 16:48:09", "content": "I agree with the accuracy discussions above.
\n\nWe do have three peaks in the accuracy histogram. One way to understand them is that accuracy of (x,y) is dependent on what technology is used in the phone for positioning. The peaks could correspond to usage of GPS, WiFi, and cellular position.
\n\nThis paper gives a good description (e.g. see Figure 3): http://www.paulzandbergen.com/PUBLICATIONS_files/Zandbergen_TGIS_2009.pdf
\n\nThe question is how to use this knowledge...
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "119902": "Hi, in training set and testing set there both exists an \"accuracy\" column, in the data description page it states that \"accuracy: location accuracy\".\r\n\r\nMy interpretation of this is that - accuracy is solely related to x,y coordinates (locations), for example a person may do facebook checkin at location (x1, y1), but in dataset it is recorded that this guy is at (x2, y2). And accuracy column is generated as a measure of difference between (x1, y1) and (x2, y2).\r\n\r\nIs the above interpretation correct? If not - then why does the data description page states \"accuracy: location accuracy\"? \r\n\r\nThanks in advance.", "119908": "I'm interpreting it as a measure of the reliability of the (x,y) values. ie the higher the accuracy, the more confidence we can have that (x2,y2) on the data are close to actual (unknown) location (x1,y1) of the user. The data page notes that the definitions are left \"intentionally vague\" so I suspect it's up to us to figure out the implications of high/low accuracy.\r\n\r\nSome of the data exploration done by others here seems to confirm the idea that high accuracy => better x,y values (as opposed to low accuracy values being better).\r\n\r\nI haven't yet been able to do anything useful with accuracy (despite having tried a few ideas), so I'm be interested to see what people end up doing with it.", "119909": "Thanks Peatle - your comment confirms my thoughts.\r\n\r\nI'm not 100% sure how accuracy is calculated, but what I'm sure now is that accuracy is ONLY related to x,y coordinates - it has nothing to do with time and place_id.\r\n\r\nIf there are different opinion welcome to chime in. Thanks again.", "124374": "I agree with the accuracy discussions above. \r\n\r\nWe do have three peaks in the accuracy histogram. One way to understand them is that accuracy of (x,y) is dependent on what technology is used in the phone for positioning. The peaks could correspond to usage of GPS, WiFi, and cellular position. \r\n\r\nThis paper gives a good description (e.g. see Figure 3): http://www.paulzandbergen.com/PUBLICATIONS_files/Zandbergen_TGIS_2009.pdf\r\n\r\nThe question is how to use this knowledge..." }, "source": "meta" }