![](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F723687%2Ffe59a1f5aa68e893b3a892db4b564d24%2FScreenshot%202024-02-16%20154906.png?generation=1708116565676590&alt=media\")
I don't understand about num_groups feature.
\nDoes this feature mean a period which measures some metrics on each case_ids?
", "messages": [ { "id": 2642486, "postDate": "2024-02-08T07:19:58.043Z", "content": "I don't understand about num_groups feature.
\nDoes this feature mean a period which measures some metrics on each case_ids?
", "rawMarkdown": "I don't understand about num_groups feature.\n\nDoes this feature mean a period which measures some metrics on each case_ids?", "votes": 14 }, { "id": 2642544, "postDate": "2024-02-08T08:24:14.750Z", "content": "Num_groupN are indices, see in Data tab
\n\n", "rawMarkdown": "Num_groupN are indices, see in Data tab\n\n>num_group1 - This is an indexing column used for the historical records of case_id in both depth=1 and depth=2 tables.\nnum_group2 - This is the second indexing column for depth=2 tables' historical records of case_id. The order of num_group1 and num_group2 is important and will be clarified in feature definitions.\nAll other raw columns in the tables serve as predictors. Their definitions can be found in the file feature_definitions.csv. For depth=0 tables, predictors can be directly used as features. However, for tables with depth>0, you may need to employ aggregation functions that will condense the historical records associated with each case_id into a single feature. In case num_group1 or num_group2 stands for person index (this is clear with predictor definitions) the zero index has special meaning. When num_groupN=0 it is the applicant (the person who applied for a loan).\n\n", "replies": [ { "id": 2642553, "postDate": "2024-02-08T08:42:32.327Z", "content": "num_group1 - This is an indexing column used for the historical records of case_id in both depth=1 and depth=2 tables.
\n
\n num_group2 - This is the second indexing column for depth=2 tables' historical records of case_id. The order of num_group1 and num_group2 is important and will be clarified in feature definitions.
\n All other raw columns in the tables serve as predictors. Their definitions can be found in the file feature_definitions.csv. For depth=0 tables, predictors can be directly used as features. However, for tables with depth>0, you may need to employ aggregation functions that will condense the historical records associated with each case_id into a single feature. In case num_group1 or num_group2 stands for person index (this is clear with predictor definitions) the zero index has special meaning. When num_groupN=0 it is the applicant (the person who applied for a loan).
Then, is structure below right?
\nA | B means that A is a source table and B is key for joining
\n-> means joining
If I want to find historical data related to a case_id,
\nbase_table | case_id -> depth1_table | case_id & num_group1 -> depth2_table
", "rawMarkdown": "Then, is structure below right?\n\nA | B means that A is a source table and B is key for joining\n-> means joining\n\nIf I want to find historical data related to a case_id,\n\nbase_table | case_id -> depth1_table | case_id & num_group1 -> depth2_table", "replies": [ { "id": 2642612, "postDate": "2024-02-08T09:33:14.017Z", "content": "Not exactly. For all the joining you need only case_id. Assume we have a table of depth=1, there will be a feature sex which stands for sex of all the people the person who applied for loan stated on the application form. Then num_group1=0 implies that it is the sex is of the person who applied for the loan, next num_group1=1 is for the first person mentioned on the application form, num_group1=2 is for second and so on. Basically num_group1 is an index for tables of depth=1. Does this answer your question?
Hmm.. Yes, helpful to understand it ! Thanks.
", "rawMarkdown": "Hmm.. Yes, helpful to understand it ! Thanks.", "votes": 1 }, { "id": 2644842, "postDate": "2024-02-09T17:56:39.390Z", "content": "Hi Daniel, can I ask you another detail?
\nIf the table has depth=1 how the num_group1 can be related to the different people and not a temporal index as stated in the first message?
\nI'm not getting this point.
We might produce a list of such features where the meaning of the index differs. I need to discuss this internally and I will get back you on Monday.
", "rawMarkdown": "We might produce a list of such features where the meaning of the index differs. I need to discuss this internally and I will get back you on Monday." }, { "id": 2650369, "postDate": "2024-02-13T12:25:14.950Z", "content": "Any news on that?
", "rawMarkdown": "Any news on that?" }, { "id": 2654996, "postDate": "2024-02-16T16:22:15.463Z", "content": "Hey Daniel,
\nI'm also a bit confused on this one. If you look at the train_applprev_2.csv file which contains all of the num_group1 and num_group2 historical associations, there's one example that's nice to look at.
\nIf you filter to case_id = 2703453, you can see num_group1 ranges from 0 to 8. So does this mean that there's up to 9 people listed on this application form?
\nAlso within depth2, when num_group1 = 2, the num_group2 values range from 0 to 3, but the conts_type_509L and credacc_cards_status_52L values are all NULL. I'm trying to understand the value these rows have if these columns are all NULL and cacccardblochreas_147M = a55475b1.
\nFor the next observation when num_group1=3, it only has 2 records for num_group2. So it's a bit hard to follow the logic.
\nSome clarity for case_id 2703453 would be super appreciated!
\nThanks,
\nMark
", "rawMarkdown": "Hey Daniel, \n\nI'm also a bit confused on this one. If you look at the train_applprev_2.csv file which contains all of the num_group1 and num_group2 historical associations, there's one example that's nice to look at. \n\nIf you filter to case_id = 2703453, you can see num_group1 ranges from 0 to 8. So does this mean that there's up to 9 people listed on this application form? \n\nAlso within depth2, when num_group1 = 2, the num_group2 values range from 0 to 3, but the conts_type_509L and credacc_cards_status_52L values are all NULL. I'm trying to understand the value these rows have if these columns are all NULL and cacccardblochreas_147M = a55475b1.\n\nFor the next observation when num_group1=3, it only has 2 records for num_group2. So it's a bit hard to follow the logic. \n\nSome clarity for case_id 2703453 would be super appreciated!\n\nThanks,\n\nMark", "votes": 3 }, { "id": 2655418, "postDate": "2024-02-16T20:51:54.710Z", "content": "It looks like the num_group1 logic is answered in other threads. So in this case, the application for case_id 2703453 had 9 previous applications. But inside of each group, I'm not sure I understand the reasoning behind the rows where conts_type_509L and credacc_cards_status_52L are both NULL? Is it because cacccardblochreas_147M maps to other values?
\n
Hi ! mark pearl! Do you solve the above problem? I get the same confusion.
", "rawMarkdown": "Hi ! mark pearl! Do you solve the above problem? I get the same confusion." } ] } ] } ], "comments": [ { "id": 2642544, "author_name": "Daniel Herman", "author_url": "", "post_date": "2024-02-08T08:24:14.750000", "content": "Num_groupN are indices, see in Data tab
\n\n", "votes": 0, "replies": [ { "id": 2642553, "author_name": "Cafelatte1", "author_url": "", "post_date": "2024-02-08T08:42:32.327000", "content": "num_group1 - This is an indexing column used for the historical records of case_id in both depth=1 and depth=2 tables.
\n
\n num_group2 - This is the second indexing column for depth=2 tables' historical records of case_id. The order of num_group1 and num_group2 is important and will be clarified in feature definitions.
\n All other raw columns in the tables serve as predictors. Their definitions can be found in the file feature_definitions.csv. For depth=0 tables, predictors can be directly used as features. However, for tables with depth>0, you may need to employ aggregation functions that will condense the historical records associated with each case_id into a single feature. In case num_group1 or num_group2 stands for person index (this is clear with predictor definitions) the zero index has special meaning. When num_groupN=0 it is the applicant (the person who applied for a loan).
Then, is structure below right?
\nA | B means that A is a source table and B is key for joining
\n-> means joining
If I want to find historical data related to a case_id,
\nbase_table | case_id -> depth1_table | case_id & num_group1 -> depth2_table
", "votes": 0, "replies": [ { "id": 2642612, "author_name": "Daniel Herman", "author_url": "", "post_date": "2024-02-08T09:33:14.017000", "content": "Not exactly. For all the joining you need only case_id. Assume we have a table of depth=1, there will be a feature sex which stands for sex of all the people the person who applied for loan stated on the application form. Then num_group1=0 implies that it is the sex is of the person who applied for the loan, next num_group1=1 is for the first person mentioned on the application form, num_group1=2 is for second and so on. Basically num_group1 is an index for tables of depth=1. Does this answer your question?
Hmm.. Yes, helpful to understand it ! Thanks.
", "votes": 1, "replies": [] }, { "id": 2644842, "author_name": "Julien94", "author_url": "", "post_date": "2024-02-09T17:56:39.390000", "content": "Hi Daniel, can I ask you another detail?
\nIf the table has depth=1 how the num_group1 can be related to the different people and not a temporal index as stated in the first message?
\nI'm not getting this point.
We might produce a list of such features where the meaning of the index differs. I need to discuss this internally and I will get back you on Monday.
", "votes": 0, "replies": [] }, { "id": 2650369, "author_name": "Julien94", "author_url": "", "post_date": "2024-02-13T12:25:14.950000", "content": "Any news on that?
", "votes": 0, "replies": [] }, { "id": 2654996, "author_name": "Mark Pearl", "author_url": "", "post_date": "2024-02-16T16:22:15.463000", "content": "Hey Daniel,
\nI'm also a bit confused on this one. If you look at the train_applprev_2.csv file which contains all of the num_group1 and num_group2 historical associations, there's one example that's nice to look at.
\nIf you filter to case_id = 2703453, you can see num_group1 ranges from 0 to 8. So does this mean that there's up to 9 people listed on this application form?
\nAlso within depth2, when num_group1 = 2, the num_group2 values range from 0 to 3, but the conts_type_509L and credacc_cards_status_52L values are all NULL. I'm trying to understand the value these rows have if these columns are all NULL and cacccardblochreas_147M = a55475b1.
\nFor the next observation when num_group1=3, it only has 2 records for num_group2. So it's a bit hard to follow the logic.
\nSome clarity for case_id 2703453 would be super appreciated!
\nThanks,
\nMark
", "votes": 3, "replies": [] }, { "id": 2655418, "author_name": "Mark Pearl", "author_url": "", "post_date": "2024-02-16T20:51:54.710000", "content": "It looks like the num_group1 logic is answered in other threads. So in this case, the application for case_id 2703453 had 9 previous applications. But inside of each group, I'm not sure I understand the reasoning behind the rows where conts_type_509L and credacc_cards_status_52L are both NULL? Is it because cacccardblochreas_147M maps to other values?
\n
Hi ! mark pearl! Do you solve the above problem? I get the same confusion.
", "votes": 0, "replies": [] } ] } ] } ], "raw_markdown_by_id": { "2642486": "I don't understand about num_groups feature.\n\nDoes this feature mean a period which measures some metrics on each case_ids?", "2642544": "Num_groupN are indices, see in Data tab\n\n>num_group1 - This is an indexing column used for the historical records of case_id in both depth=1 and depth=2 tables.\nnum_group2 - This is the second indexing column for depth=2 tables' historical records of case_id. The order of num_group1 and num_group2 is important and will be clarified in feature definitions.\nAll other raw columns in the tables serve as predictors. Their definitions can be found in the file feature_definitions.csv. For depth=0 tables, predictors can be directly used as features. However, for tables with depth>0, you may need to employ aggregation functions that will condense the historical records associated with each case_id into a single feature. In case num_group1 or num_group2 stands for person index (this is clear with predictor definitions) the zero index has special meaning. When num_groupN=0 it is the applicant (the person who applied for a loan).\n\n" } }