{ "id": 23037, "title": " L0_S0_D1 <===> L0_S0_F0", "url": "/competitions/bosch-production-line-performance/discussion/23037", "author_name": "", "post_date": "2016-08-22T18:02:00.327Z", "votes": 10, "comment_count": 10, "views": 2260, "content": "

"...value of L0_S0_D1 is the time at which L0_S0_F0 was taken." <-- is there any reason for this "+1 " naming convention? Why not the more logical... L0_S0_D0 === L0_S0_F0 ? me puzzled.

", "messages": [ { "id": "132114", "postDate": "08/22/2016 18:02:00", "content": "

"...value of L0_S0_D1 is the time at which L0_S0_F0 was taken." <-- is there any reason for this "+1 " naming convention? Why not the more logical... L0_S0_D0 === L0_S0_F0 ? me puzzled.

", "rawMarkdown": "\"...value of L0_S0_D1 is the time at which L0_S0_F0 was taken.\" <-- is there any reason for this \"+1 \" naming convention? Why not the more logical... L0_S0_D0 === L0_S0_F0 ? me puzzled.", "votes": null }, { "id": "132159", "postDate": "08/23/2016 04:55:00", "content": "

Jose- the same thing crossed my mind. On my to-do list is a quick script to standardize all the headers. Seems like that would make it easier to match columns based on the suffix.

", "rawMarkdown": "Jose- the same thing crossed my mind. On my to-do list is a quick script to standardize all the headers. Seems like that would make it easier to match columns based on the suffix.", "votes": null }, { "id": "132202", "postDate": "08/23/2016 14:22:02", "content": "

Sorry for any added confusion! The features were originally in one big file and located next to each other (L0_S0_F0, L0_S0_D1...). We decided to break up the file make handling them easier.

", "rawMarkdown": "Sorry for any added confusion! The features were originally in one big file and located next to each other (L0_S0_F0, L0_S0_D1...). We decided to break up the file make handling them easier.", "votes": null }, { "id": "132216", "postDate": "08/23/2016 16:22:25", "content": "

@William -

\n\n

It's not clear to me how your response answers the original question. How does the "+1" numbering have anything to do with whether or not the features were originally in one file? Thx.

", "rawMarkdown": "William -\r\n\r\nIt's not clear to me how your response answers the original question. How does the \"+1\" numbering have anything to do with whether or not the features were originally in one file? Thx.", "votes": null }, { "id": "132218", "postDate": "08/23/2016 17:03:48", "content": "

@inversion was just explaining that the features were originally literally next to each other in one file (feature, date, feature, date, etc.) and we (Kaggle) separated them into different files. The names were kept as-is to match Bosch's own naming convention.

", "rawMarkdown": "inversion was just explaining that the features were originally literally next to each other in one file (feature, date, feature, date, etc.) and we (Kaggle) separated them into different files. The names were kept as-is to match Bosch's own naming convention.", "votes": null }, { "id": "132356", "postDate": "08/24/2016 20:09:04", "content": "

All Good. :) For clarity I would rephrase...

\n\n
\n

Each date column ends in a number that corresponds to the previous feature number. E.g. the value of L0_S0_D1 is the time at which L0_S0_F0 was taken.

\n
\n\n

for

\n\n
\n

Naming convention of column-names of the test_date.csv: L0_S0_D(K+1) contains the sampling timestamps of L0_S0_F(K) where K is a feature_id and L0_S0_F(K) is a column name in test_numeric or test_categoric.csv

\n
", "rawMarkdown": "All Good. :) For clarity I would rephrase...\r\n\r\n> Each date column ends in a number that corresponds to the previous feature number. E.g. the value of L0_S0_D1 is the time at which L0_S0_F0 was taken.\r\n\r\nfor\r\n\r\n>Naming convention of column-names of the test_date.csv: L0_S0_D(K+1) contains the sampling timestamps of L0_S0_F(K) where K is a feature_id and L0_S0_F(K) is a column name in test_numeric or test_categoric.csv", "votes": null }, { "id": "132682", "postDate": "08/28/2016 11:16:25", "content": "

Out of context, but just FYI, I assume there is an error in the column names: L3_S36_F3941 L3_S37_D3942. The date column corresponding to the station S36's feature L3_S36_F3941 has got station mentioned as S37.

", "rawMarkdown": "Out of context, but just FYI, I assume there is an error in the column names: L3_S36_F3941 L3_S37_D3942. The date column corresponding to the station S36's feature L3_S36_F3941 has got station mentioned as S37.", "votes": null }, { "id": "132854", "postDate": "08/29/2016 20:15:25", "content": "

There are D columns where there is no corresponding F column, either the lowest numbered column is a D or there are D columns following each other.

\n\n
L0 S7 D140 D141\nL0 S8 D148\nL0 S19 D457 D458\nL0 S20 D465\nL1 S24 D1826 D1833\nL1 S25 D2235 D2236 D2445 D2446 D2453 D2502 D2503 D2728 D2729 D2736 D2785 D2786 D3011 D3012 D3019\nL3 S34 D3875\nL3 S37 D3942 D3943\nL3 S41 D4010 D4013 D4022 D4025\nL3 S48 D4195\nL3 S51 D4255\n
", "rawMarkdown": "There are D columns where there is no corresponding F column, either the lowest numbered column is a D or there are D columns following each other.\r\n\r\n L0 S7 D140 D141\r\n L0 S8 D148\r\n L0 S19 D457 D458\r\n L0 S20 D465\r\n L1 S24 D1826 D1833\r\n L1 S25 D2235 D2236 D2445 D2446 D2453 D2502 D2503 D2728 D2729 D2736 D2785 D2786 D3011 D3012 D3019\r\n L3 S34 D3875\r\n L3 S37 D3942 D3943\r\n L3 S41 D4010 D4013 D4022 D4025\r\n L3 S48 D4195\r\n L3 S51 D4255", "votes": null }, { "id": "132888", "postDate": "08/30/2016 08:29:53", "content": "

It looks like the numbers of the features are a single sequence across the numeric, categorical, and date rows. Here are the first 20 feature names for numeric, categorical, and date, respectively:

\n\n
>>> list(zip(numnames, catnames, datenames))[1:20]\n[('L0_S0_F0', 'L0_S1_F25', 'L0_S0_D1'),    \n('L0_S0_F2', 'L0_S1_F27', 'L0_S0_D3'),     \n('L0_S0_F4', 'L0_S1_F29', 'L0_S0_D5'),    \n('L0_S0_F6', 'L0_S1_F31', 'L0_S0_D7'),\n('L0_S0_F8', 'L0_S2_F33', 'L0_S0_D9'),\n('L0_S0_F10', 'L0_S2_F35', 'L0_S0_D11'),\n('L0_S0_F12', 'L0_S2_F37', 'L0_S0_D13'),\n('L0_S0_F14', 'L0_S2_F39', 'L0_S0_D15'),\n('L0_S0_F16', 'L0_S2_F41', 'L0_S0_D17'),\n('L0_S0_F18', 'L0_S2_F43', 'L0_S0_D19'),\n('L0_S0_F20', 'L0_S2_F45', 'L0_S0_D21'),\n('L0_S0_F22', 'L0_S2_F47', 'L0_S0_D23'),\n('L0_S1_F24', 'L0_S2_F49', 'L0_S1_D26'),\n('L0_S1_F28', 'L0_S2_F51', 'L0_S1_D30'),\n('L0_S2_F32', 'L0_S2_F53', 'L0_S2_D34'),\n('L0_S2_F36', 'L0_S2_F55', 'L0_S2_D38'),\n('L0_S2_F40', 'L0_S2_F57', 'L0_S2_D42'),\n('L0_S2_F44', 'L0_S2_F59', 'L0_S2_D46'),\n('L0_S2_F48', 'L0_S2_F61', 'L0_S2_D50')]    \n
\n\n

There is just one numeric sequence overall: F0 is numeric, D1 is a date...F24 is categorical, F25 is numeric, D26 is a date, F27 is categorical, F28 is numeric...

\n\n

Also, I'm getting that there are 1156 date entries, 968 numeric and 2140 categorical, so there isn't a 1-1 relationship between the timestamps and the features. So what I want to know is: which features match up with which dates?

", "rawMarkdown": "It looks like the numbers of the features are a single sequence across the numeric, categorical, and date rows. Here are the first 20 feature names for numeric, categorical, and date, respectively:\r\n\r\n >>> list(zip(numnames, catnames, datenames))[1:20]\r\n [('L0_S0_F0', 'L0_S1_F25', 'L0_S0_D1'), \r\n ('L0_S0_F2', 'L0_S1_F27', 'L0_S0_D3'), \r\n ('L0_S0_F4', 'L0_S1_F29', 'L0_S0_D5'), \r\n ('L0_S0_F6', 'L0_S1_F31', 'L0_S0_D7'),\r\n ('L0_S0_F8', 'L0_S2_F33', 'L0_S0_D9'),\r\n ('L0_S0_F10', 'L0_S2_F35', 'L0_S0_D11'),\r\n ('L0_S0_F12', 'L0_S2_F37', 'L0_S0_D13'),\r\n ('L0_S0_F14', 'L0_S2_F39', 'L0_S0_D15'),\r\n ('L0_S0_F16', 'L0_S2_F41', 'L0_S0_D17'),\r\n ('L0_S0_F18', 'L0_S2_F43', 'L0_S0_D19'),\r\n ('L0_S0_F20', 'L0_S2_F45', 'L0_S0_D21'),\r\n ('L0_S0_F22', 'L0_S2_F47', 'L0_S0_D23'),\r\n ('L0_S1_F24', 'L0_S2_F49', 'L0_S1_D26'),\r\n ('L0_S1_F28', 'L0_S2_F51', 'L0_S1_D30'),\r\n ('L0_S2_F32', 'L0_S2_F53', 'L0_S2_D34'),\r\n ('L0_S2_F36', 'L0_S2_F55', 'L0_S2_D38'),\r\n ('L0_S2_F40', 'L0_S2_F57', 'L0_S2_D42'),\r\n ('L0_S2_F44', 'L0_S2_F59', 'L0_S2_D46'),\r\n ('L0_S2_F48', 'L0_S2_F61', 'L0_S2_D50')] \r\n\r\nThere is just one numeric sequence overall: F0 is numeric, D1 is a date...F24 is categorical, F25 is numeric, D26 is a date, F27 is categorical, F28 is numeric... \r\n\r\nAlso, I'm getting that there are 1156 date entries, 968 numeric and 2140 categorical, so there isn't a 1-1 relationship between the timestamps and the features. So what I want to know is: which features match up with which dates?", "votes": null }, { "id": "132938", "postDate": "08/30/2016 13:47:35", "content": "

Yo @william Do you have the data in tall form?

\n\n

[ feature, time stamp, value ] \n .\n .\n .

\n\n

?

", "rawMarkdown": "Yo @william Do you have the data in tall form?\r\n\r\n[ feature, time stamp, value ] \r\n .\r\n .\r\n .\r\n\r\n?", "votes": null }, { "id": "132945", "postDate": "08/30/2016 14:42:23", "content": "

[quote=David Thaler;132888]

\n\n

Also, I'm getting that there are 1156 date entries, 968 numeric and 2140 categorical, so there isn't a 1-1 relationship between the timestamps and the features. So what I want to know is: which features match up with which dates?

\n\n

[/quote]

\n\n

There are many instances where you will have multiple features before a timestamp column, and (particularly in stations 24 and 25) multiple instances of timestamp columns with no features in-between. Also, there are numerous instances of clusters of columns of the same type of feature (either categorical or numeric). A simple string substitution for the categorical headers from "_F" to "_C" can help to illuminate this.

\n\n

I am working on the following assumptions about this dataset:

\n\n

All features are measurements of some kind, but not necessarily anything to do with defects specifically

\n\n

All measurements are automated through sensors with datalogging capabilities

\n\n

Not all sensors send the data with timestamps OR clusters of sensors use the same datalogging connection points, which may explain consecutive features before individual timestamps

\n\n

My personal belief is that for most stations, the timestamp is the the time the part is "checked-in" to that station, and not the exact time of measurement. If true, the timestamps carry less value than originally intended.

", "rawMarkdown": "[quote=David Thaler;132888]\r\n\r\nAlso, I'm getting that there are 1156 date entries, 968 numeric and 2140 categorical, so there isn't a 1-1 relationship between the timestamps and the features. So what I want to know is: which features match up with which dates?\r\n\r\n[/quote]\r\n\r\nThere are many instances where you will have multiple features before a timestamp column, and (particularly in stations 24 and 25) multiple instances of timestamp columns with no features in-between. Also, there are numerous instances of clusters of columns of the same type of feature (either categorical or numeric). A simple string substitution for the categorical headers from \"_F\" to \"_C\" can help to illuminate this.\r\n\r\nI am working on the following assumptions about this dataset:\r\n\r\nAll features are measurements of some kind, but not necessarily anything to do with defects specifically\r\n\r\nAll measurements are automated through sensors with datalogging capabilities\r\n\r\nNot all sensors send the data with timestamps OR clusters of sensors use the same datalogging connection points, which may explain consecutive features before individual timestamps\r\n\r\nMy personal belief is that for most stations, the timestamp is the the time the part is \"checked-in\" to that station, and not the exact time of measurement. If true, the timestamps carry less value than originally intended.", "votes": null } ], "comments": [ { "id": 132159, "author_name": "jpmiller", "author_url": "", "post_date": "08/23/2016 04:55:00", "content": "

Jose- the same thing crossed my mind. On my to-do list is a quick script to standardize all the headers. Seems like that would make it easier to match columns based on the suffix.

", "votes": null, "replies": [] }, { "id": 132202, "author_name": "wcukierski", "author_url": "", "post_date": "08/23/2016 14:22:02", "content": "

Sorry for any added confusion! The features were originally in one big file and located next to each other (L0_S0_F0, L0_S0_D1...). We decided to break up the file make handling them easier.

", "votes": null, "replies": [] }, { "id": 132216, "author_name": "inversion", "author_url": "", "post_date": "08/23/2016 16:22:25", "content": "

@William -

\n\n

It's not clear to me how your response answers the original question. How does the "+1" numbering have anything to do with whether or not the features were originally in one file? Thx.

", "votes": null, "replies": [] }, { "id": 132218, "author_name": "wcukierski", "author_url": "", "post_date": "08/23/2016 17:03:48", "content": "

@inversion was just explaining that the features were originally literally next to each other in one file (feature, date, feature, date, etc.) and we (Kaggle) separated them into different files. The names were kept as-is to match Bosch's own naming convention.

", "votes": null, "replies": [] }, { "id": 132356, "author_name": "harriken", "author_url": "", "post_date": "08/24/2016 20:09:04", "content": "

All Good. :) For clarity I would rephrase...

\n\n
\n

Each date column ends in a number that corresponds to the previous feature number. E.g. the value of L0_S0_D1 is the time at which L0_S0_F0 was taken.

\n
\n\n

for

\n\n
\n

Naming convention of column-names of the test_date.csv: L0_S0_D(K+1) contains the sampling timestamps of L0_S0_F(K) where K is a feature_id and L0_S0_F(K) is a column name in test_numeric or test_categoric.csv

\n
", "votes": null, "replies": [] }, { "id": 132682, "author_name": "manikazad", "author_url": "", "post_date": "08/28/2016 11:16:25", "content": "

Out of context, but just FYI, I assume there is an error in the column names: L3_S36_F3941 L3_S37_D3942. The date column corresponding to the station S36's feature L3_S36_F3941 has got station mentioned as S37.

", "votes": null, "replies": [] }, { "id": 132854, "author_name": "attilababo", "author_url": "", "post_date": "08/29/2016 20:15:25", "content": "

There are D columns where there is no corresponding F column, either the lowest numbered column is a D or there are D columns following each other.

\n\n
L0 S7 D140 D141\nL0 S8 D148\nL0 S19 D457 D458\nL0 S20 D465\nL1 S24 D1826 D1833\nL1 S25 D2235 D2236 D2445 D2446 D2453 D2502 D2503 D2728 D2729 D2736 D2785 D2786 D3011 D3012 D3019\nL3 S34 D3875\nL3 S37 D3942 D3943\nL3 S41 D4010 D4013 D4022 D4025\nL3 S48 D4195\nL3 S51 D4255\n
", "votes": null, "replies": [] }, { "id": 132888, "author_name": "davidthaler", "author_url": "", "post_date": "08/30/2016 08:29:53", "content": "

It looks like the numbers of the features are a single sequence across the numeric, categorical, and date rows. Here are the first 20 feature names for numeric, categorical, and date, respectively:

\n\n
>>> list(zip(numnames, catnames, datenames))[1:20]\n[('L0_S0_F0', 'L0_S1_F25', 'L0_S0_D1'),    \n('L0_S0_F2', 'L0_S1_F27', 'L0_S0_D3'),     \n('L0_S0_F4', 'L0_S1_F29', 'L0_S0_D5'),    \n('L0_S0_F6', 'L0_S1_F31', 'L0_S0_D7'),\n('L0_S0_F8', 'L0_S2_F33', 'L0_S0_D9'),\n('L0_S0_F10', 'L0_S2_F35', 'L0_S0_D11'),\n('L0_S0_F12', 'L0_S2_F37', 'L0_S0_D13'),\n('L0_S0_F14', 'L0_S2_F39', 'L0_S0_D15'),\n('L0_S0_F16', 'L0_S2_F41', 'L0_S0_D17'),\n('L0_S0_F18', 'L0_S2_F43', 'L0_S0_D19'),\n('L0_S0_F20', 'L0_S2_F45', 'L0_S0_D21'),\n('L0_S0_F22', 'L0_S2_F47', 'L0_S0_D23'),\n('L0_S1_F24', 'L0_S2_F49', 'L0_S1_D26'),\n('L0_S1_F28', 'L0_S2_F51', 'L0_S1_D30'),\n('L0_S2_F32', 'L0_S2_F53', 'L0_S2_D34'),\n('L0_S2_F36', 'L0_S2_F55', 'L0_S2_D38'),\n('L0_S2_F40', 'L0_S2_F57', 'L0_S2_D42'),\n('L0_S2_F44', 'L0_S2_F59', 'L0_S2_D46'),\n('L0_S2_F48', 'L0_S2_F61', 'L0_S2_D50')]    \n
\n\n

There is just one numeric sequence overall: F0 is numeric, D1 is a date...F24 is categorical, F25 is numeric, D26 is a date, F27 is categorical, F28 is numeric...

\n\n

Also, I'm getting that there are 1156 date entries, 968 numeric and 2140 categorical, so there isn't a 1-1 relationship between the timestamps and the features. So what I want to know is: which features match up with which dates?

", "votes": null, "replies": [] }, { "id": 132938, "author_name": "harriken", "author_url": "", "post_date": "08/30/2016 13:47:35", "content": "

Yo @william Do you have the data in tall form?

\n\n

[ feature, time stamp, value ] \n .\n .\n .

\n\n

?

", "votes": null, "replies": [] }, { "id": 132945, "author_name": "taffeylewis", "author_url": "", "post_date": "08/30/2016 14:42:23", "content": "

[quote=David Thaler;132888]

\n\n

Also, I'm getting that there are 1156 date entries, 968 numeric and 2140 categorical, so there isn't a 1-1 relationship between the timestamps and the features. So what I want to know is: which features match up with which dates?

\n\n

[/quote]

\n\n

There are many instances where you will have multiple features before a timestamp column, and (particularly in stations 24 and 25) multiple instances of timestamp columns with no features in-between. Also, there are numerous instances of clusters of columns of the same type of feature (either categorical or numeric). A simple string substitution for the categorical headers from "_F" to "_C" can help to illuminate this.

\n\n

I am working on the following assumptions about this dataset:

\n\n

All features are measurements of some kind, but not necessarily anything to do with defects specifically

\n\n

All measurements are automated through sensors with datalogging capabilities

\n\n

Not all sensors send the data with timestamps OR clusters of sensors use the same datalogging connection points, which may explain consecutive features before individual timestamps

\n\n

My personal belief is that for most stations, the timestamp is the the time the part is "checked-in" to that station, and not the exact time of measurement. If true, the timestamps carry less value than originally intended.

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "132114": "\"...value of L0_S0_D1 is the time at which L0_S0_F0 was taken.\" <-- is there any reason for this \"+1 \" naming convention? Why not the more logical... L0_S0_D0 === L0_S0_F0 ? me puzzled.", "132159": "Jose- the same thing crossed my mind. On my to-do list is a quick script to standardize all the headers. Seems like that would make it easier to match columns based on the suffix.", "132202": "Sorry for any added confusion! The features were originally in one big file and located next to each other (L0_S0_F0, L0_S0_D1...). We decided to break up the file make handling them easier.", "132216": "William -\r\n\r\nIt's not clear to me how your response answers the original question. How does the \"+1\" numbering have anything to do with whether or not the features were originally in one file? Thx.", "132218": "inversion was just explaining that the features were originally literally next to each other in one file (feature, date, feature, date, etc.) and we (Kaggle) separated them into different files. The names were kept as-is to match Bosch's own naming convention.", "132356": "All Good. :) For clarity I would rephrase...\r\n\r\n> Each date column ends in a number that corresponds to the previous feature number. E.g. the value of L0_S0_D1 is the time at which L0_S0_F0 was taken.\r\n\r\nfor\r\n\r\n>Naming convention of column-names of the test_date.csv: L0_S0_D(K+1) contains the sampling timestamps of L0_S0_F(K) where K is a feature_id and L0_S0_F(K) is a column name in test_numeric or test_categoric.csv", "132682": "Out of context, but just FYI, I assume there is an error in the column names: L3_S36_F3941 L3_S37_D3942. The date column corresponding to the station S36's feature L3_S36_F3941 has got station mentioned as S37.", "132854": "There are D columns where there is no corresponding F column, either the lowest numbered column is a D or there are D columns following each other.\r\n\r\n L0 S7 D140 D141\r\n L0 S8 D148\r\n L0 S19 D457 D458\r\n L0 S20 D465\r\n L1 S24 D1826 D1833\r\n L1 S25 D2235 D2236 D2445 D2446 D2453 D2502 D2503 D2728 D2729 D2736 D2785 D2786 D3011 D3012 D3019\r\n L3 S34 D3875\r\n L3 S37 D3942 D3943\r\n L3 S41 D4010 D4013 D4022 D4025\r\n L3 S48 D4195\r\n L3 S51 D4255", "132888": "It looks like the numbers of the features are a single sequence across the numeric, categorical, and date rows. Here are the first 20 feature names for numeric, categorical, and date, respectively:\r\n\r\n >>> list(zip(numnames, catnames, datenames))[1:20]\r\n [('L0_S0_F0', 'L0_S1_F25', 'L0_S0_D1'), \r\n ('L0_S0_F2', 'L0_S1_F27', 'L0_S0_D3'), \r\n ('L0_S0_F4', 'L0_S1_F29', 'L0_S0_D5'), \r\n ('L0_S0_F6', 'L0_S1_F31', 'L0_S0_D7'),\r\n ('L0_S0_F8', 'L0_S2_F33', 'L0_S0_D9'),\r\n ('L0_S0_F10', 'L0_S2_F35', 'L0_S0_D11'),\r\n ('L0_S0_F12', 'L0_S2_F37', 'L0_S0_D13'),\r\n ('L0_S0_F14', 'L0_S2_F39', 'L0_S0_D15'),\r\n ('L0_S0_F16', 'L0_S2_F41', 'L0_S0_D17'),\r\n ('L0_S0_F18', 'L0_S2_F43', 'L0_S0_D19'),\r\n ('L0_S0_F20', 'L0_S2_F45', 'L0_S0_D21'),\r\n ('L0_S0_F22', 'L0_S2_F47', 'L0_S0_D23'),\r\n ('L0_S1_F24', 'L0_S2_F49', 'L0_S1_D26'),\r\n ('L0_S1_F28', 'L0_S2_F51', 'L0_S1_D30'),\r\n ('L0_S2_F32', 'L0_S2_F53', 'L0_S2_D34'),\r\n ('L0_S2_F36', 'L0_S2_F55', 'L0_S2_D38'),\r\n ('L0_S2_F40', 'L0_S2_F57', 'L0_S2_D42'),\r\n ('L0_S2_F44', 'L0_S2_F59', 'L0_S2_D46'),\r\n ('L0_S2_F48', 'L0_S2_F61', 'L0_S2_D50')] \r\n\r\nThere is just one numeric sequence overall: F0 is numeric, D1 is a date...F24 is categorical, F25 is numeric, D26 is a date, F27 is categorical, F28 is numeric... \r\n\r\nAlso, I'm getting that there are 1156 date entries, 968 numeric and 2140 categorical, so there isn't a 1-1 relationship between the timestamps and the features. So what I want to know is: which features match up with which dates?", "132938": "Yo @william Do you have the data in tall form?\r\n\r\n[ feature, time stamp, value ] \r\n .\r\n .\r\n .\r\n\r\n?", "132945": "[quote=David Thaler;132888]\r\n\r\nAlso, I'm getting that there are 1156 date entries, 968 numeric and 2140 categorical, so there isn't a 1-1 relationship between the timestamps and the features. So what I want to know is: which features match up with which dates?\r\n\r\n[/quote]\r\n\r\nThere are many instances where you will have multiple features before a timestamp column, and (particularly in stations 24 and 25) multiple instances of timestamp columns with no features in-between. Also, there are numerous instances of clusters of columns of the same type of feature (either categorical or numeric). A simple string substitution for the categorical headers from \"_F\" to \"_C\" can help to illuminate this.\r\n\r\nI am working on the following assumptions about this dataset:\r\n\r\nAll features are measurements of some kind, but not necessarily anything to do with defects specifically\r\n\r\nAll measurements are automated through sensors with datalogging capabilities\r\n\r\nNot all sensors send the data with timestamps OR clusters of sensors use the same datalogging connection points, which may explain consecutive features before individual timestamps\r\n\r\nMy personal belief is that for most stations, the timestamp is the the time the part is \"checked-in\" to that station, and not the exact time of measurement. If true, the timestamps carry less value than originally intended." }, "source": "meta" }