My project group is confused about which ID numbers ('msno') are supposed to be included in the test set. The submissions page says you must submit a file with exactly 907,471 prediction rows, but we can't figure out where that number comes from.
\n\nWe thought some of the members in the test set might have ended up churning, so we checked if removing them from test set would get us the correct number of members. Cross-checking with train.csv and train_v2.csv, we identified 62,262 unique members who were in the test set but had churned. Removing these takes us down to 926,339 members, which is not the correct number.
\n\nAlso, submitting a random selection of 907,471 results in an error that some specific IDs are missing.
\n\nDoes anyone know what we're doing wrong?
", "messages": [ { "id": "243502", "postDate": "11/14/2017 07:36:58", "content": "My project group is confused about which ID numbers ('msno') are supposed to be included in the test set. The submissions page says you must submit a file with exactly 907,471 prediction rows, but we can't figure out where that number comes from.
\n\nWe thought some of the members in the test set might have ended up churning, so we checked if removing them from test set would get us the correct number of members. Cross-checking with train.csv and train_v2.csv, we identified 62,262 unique members who were in the test set but had churned. Removing these takes us down to 926,339 members, which is not the correct number.
\n\nAlso, submitting a random selection of 907,471 results in an error that some specific IDs are missing.
\n\nDoes anyone know what we're doing wrong?
", "rawMarkdown": "My project group is confused about which ID numbers ('msno') are supposed to be included in the test set. The submissions page says you must submit a file with *exactly* 907,471 prediction rows, but we can't figure out where that number comes from.\n\n- sample_submissions_zero.csv has 970,960 rows\n- sample_submissions_v2 has 151,465 rows\n- If you combine the 2 sample submission files and remove duplicates, there are 988,601 rows\n\nWe thought some of the members in the test set might have ended up churning, so we checked if removing them from test set would get us the correct number of members. Cross-checking with train.csv and train_v2.csv, we identified 62,262 unique members who were in the test set but had churned. Removing these takes us down to 926,339 members, which is not the correct number.\n\nAlso, submitting a random selection of 907,471 results in an error that some specific IDs are missing.\n\nDoes anyone know what we're doing wrong?", "votes": null }, { "id": "243542", "postDate": "11/14/2017 09:53:15", "content": "the sample_submission_zero is no longer valid. Please make prediction for the the msno listed in the sample_submissions_v2.
", "rawMarkdown": "the sample_submission_zero is no longer valid. Please make prediction for the the msno listed in the sample_submissions_v2.", "votes": null }, { "id": "243673", "postDate": "11/14/2017 17:03:29", "content": "Hi Arden, thanks for getting back to us. The problem we have is that the submission page does not allow us to submit a file unless it is exactly 907,471 prediction rows. The sample_submissions_v2 file has 151,465 msno's. So when we try to submit predictions for only the msno's in v2, we get an error that the file doesn't have the correct number of rows.
", "rawMarkdown": "Hi Arden, thanks for getting back to us. The problem we have is that the submission page does not allow us to submit a file unless it is *exactly* 907,471 prediction rows. The sample_submissions_v2 file has 151,465 msno's. So when we try to submit predictions for only the msno's in v2, we get an error that the file doesn't have the correct number of rows.", "votes": null }, { "id": "243691", "postDate": "11/14/2017 17:41:46", "content": "Have you tried re-downloading and extracting sample_submissions_v2 again? The number of lines in the file inside the resulting data/churn_comp_refresh/ directory that is created should indeed have 907,471 lines.
", "rawMarkdown": "Have you tried re-downloading and extracting sample_submissions_v2 again? The number of lines in the file inside the resulting data/churn_comp_refresh/ directory that is created should indeed have 907,471 lines.", "votes": null }, { "id": "243692", "postDate": "11/14/2017 17:47:44", "content": "Ahh thank you, that was the issue! I need to be keeping up with the data files better
", "rawMarkdown": "Ahh thank you, that was the issue! I need to be keeping up with the data files better", "votes": null } ], "comments": [ { "id": 243542, "author_name": "ardenkkbox", "author_url": "", "post_date": "11/14/2017 09:53:15", "content": "the sample_submission_zero is no longer valid. Please make prediction for the the msno listed in the sample_submissions_v2.
", "votes": null, "replies": [ { "id": 243673, "author_name": "carmen16", "author_url": "", "post_date": "11/14/2017 17:03:29", "content": "Hi Arden, thanks for getting back to us. The problem we have is that the submission page does not allow us to submit a file unless it is exactly 907,471 prediction rows. The sample_submissions_v2 file has 151,465 msno's. So when we try to submit predictions for only the msno's in v2, we get an error that the file doesn't have the correct number of rows.
", "votes": null, "replies": [] }, { "id": 243691, "author_name": "marianaoliveira", "author_url": "", "post_date": "11/14/2017 17:41:46", "content": "Have you tried re-downloading and extracting sample_submissions_v2 again? The number of lines in the file inside the resulting data/churn_comp_refresh/ directory that is created should indeed have 907,471 lines.
", "votes": null, "replies": [] }, { "id": 243692, "author_name": "carmen16", "author_url": "", "post_date": "11/14/2017 17:47:44", "content": "Ahh thank you, that was the issue! I need to be keeping up with the data files better
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "243502": "My project group is confused about which ID numbers ('msno') are supposed to be included in the test set. The submissions page says you must submit a file with *exactly* 907,471 prediction rows, but we can't figure out where that number comes from.\n\n- sample_submissions_zero.csv has 970,960 rows\n- sample_submissions_v2 has 151,465 rows\n- If you combine the 2 sample submission files and remove duplicates, there are 988,601 rows\n\nWe thought some of the members in the test set might have ended up churning, so we checked if removing them from test set would get us the correct number of members. Cross-checking with train.csv and train_v2.csv, we identified 62,262 unique members who were in the test set but had churned. Removing these takes us down to 926,339 members, which is not the correct number.\n\nAlso, submitting a random selection of 907,471 results in an error that some specific IDs are missing.\n\nDoes anyone know what we're doing wrong?", "243542": "the sample_submission_zero is no longer valid. Please make prediction for the the msno listed in the sample_submissions_v2.", "243673": "Hi Arden, thanks for getting back to us. The problem we have is that the submission page does not allow us to submit a file unless it is *exactly* 907,471 prediction rows. The sample_submissions_v2 file has 151,465 msno's. So when we try to submit predictions for only the msno's in v2, we get an error that the file doesn't have the correct number of rows.", "243691": "Have you tried re-downloading and extracting sample_submissions_v2 again? The number of lines in the file inside the resulting data/churn_comp_refresh/ directory that is created should indeed have 907,471 lines.", "243692": "Ahh thank you, that was the issue! I need to be keeping up with the data files better" }, "source": "meta" }