The definition the competition gives us is: \"item_seq_number - Ad sequential number for user.\". That seems very vague to me, is it a numerical or categorical feature?
", "messages": [ { "id": "326595", "postDate": "05/10/2018 01:57:57", "content": "The definition the competition gives us is: \"item_seq_number - Ad sequential number for user.\". That seems very vague to me, is it a numerical or categorical feature?
", "rawMarkdown": "The definition the competition gives us is: \"item_seq_number - Ad sequential number for user.\". That seems very vague to me, is it a numerical or categorical feature?", "votes": null }, { "id": "326602", "postDate": "05/10/2018 02:13:38", "content": "I agree it is vague and clarification on this variable, and many other variables (including the target!) would be helpful.
\n\nFrom looking at the variable, my best guess is that an item_seq_number of X means that is the Xth ad the user posted.
\n\nI infer this because:\n(a) item_seq_number is higher as time is higher and\n(b) mean item_seq_number is much higher for user_types I expect post more -- \"Company\" (mean 646.2) and \"Shop\" (mean 10633) is much higher than for \"Private\" user_type (35.3).
", "rawMarkdown": "I agree it is vague and clarification on this variable, and many other variables (including the target!) would be helpful.\n\nFrom looking at the variable, my best guess is that an item_seq_number of X means that is the Xth ad the user posted.\n\nI infer this because:\n(a) item_seq_number is higher as time is higher and\n(b) mean item_seq_number is much higher for user_types I expect post more -- \"Company\" (mean 646.2) and \"Shop\" (mean 10633) is much higher than for \"Private\" user_type (35.3).", "votes": null }, { "id": "326605", "postDate": "05/10/2018 02:26:38", "content": "That would make sense if that is what it is, and why it could work as either categorical or numerical
", "rawMarkdown": "That would make sense if that is what it is, and why it could work as either categorical or numerical", "votes": null }, { "id": "326607", "postDate": "05/10/2018 02:28:38", "content": "If my interpretation is correct, it is technically an \"ordinal\". So far I've been treating it as numeric.
", "rawMarkdown": "If my interpretation is correct, it is technically an \"ordinal\". So far I've been treating it as numeric.", "votes": null }, { "id": "326663", "postDate": "05/10/2018 05:51:58", "content": "Agree ! with Peter that this variable can be the sequence of images uploaded by the user but an addition to it is that its since registration . Most of the user item_seq_number starts from 1 . Some which has item_sequence_number more than 1 means that , first image loaded by them is in the past and is not present in train set time duration . I believe 72385 users (i.e. ~ 9% ) has a first time post on the website in train set duration.
\n\n(train.groupby('user_id')['item_seq_number'].max() <= 1).value_counts()
\n\nCorrect me if my understanding is wrong .
", "rawMarkdown": "Agree ! with Peter that this variable can be the sequence of images uploaded by the user but an addition to it is that its since registration . Most of the user item_seq_number starts from 1 . Some which has item_sequence_number more than 1 means that , first image loaded by them is in the past and is not present in train set time duration . I believe 72385 users (i.e. ~ 9% ) has a first time post on the website in train set duration. \n\n(train.groupby('user_id')['item_seq_number'].max() <= 1).value_counts()\n\nCorrect me if my understanding is wrong .", "votes": null }, { "id": "326716", "postDate": "05/10/2018 08:16:03", "content": "Its something like AD ID system, check this out http://www.ad-id.org/how-it-works/ad-id-structure
", "rawMarkdown": "Its something like AD ID system, check this out http://www.ad-id.org/how-it-works/ad-id-structure", "votes": null }, { "id": "329991", "postDate": "05/17/2018 19:02:10", "content": "@ Peter \nI am treating this as a numeric myself. IS that how you are doing it for your current LB position?
", "rawMarkdown": "Peter \nI am treating this as a numeric myself. IS that how you are doing it for your current LB position?", "votes": null }, { "id": "329992", "postDate": "05/17/2018 19:07:36", "content": "Found another discussion, that is talking about interpreting this variable. \nThe user talks about how this can be, \"A Sequential number in which the specific item_id (unique AD) will be displayed to the same user_id across multiple devices\"
\n\nDiscussion link is here How to interpret item_seq_number
", "rawMarkdown": "Found another discussion, that is talking about interpreting this variable. \nThe user talks about how this can be, \"A Sequential number in which the specific item_id (unique AD) will be displayed to the same user_id across multiple devices\"\n\nDiscussion link is here [How to interpret item_seq_number][1]\n\n\n [1]: https://www.kaggle.com/c/avito-demand-prediction/discussion/56693", "votes": null } ], "comments": [ { "id": 326602, "author_name": "peterhurford", "author_url": "", "post_date": "05/10/2018 02:13:38", "content": "I agree it is vague and clarification on this variable, and many other variables (including the target!) would be helpful.
\n\nFrom looking at the variable, my best guess is that an item_seq_number of X means that is the Xth ad the user posted.
\n\nI infer this because:\n(a) item_seq_number is higher as time is higher and\n(b) mean item_seq_number is much higher for user_types I expect post more -- \"Company\" (mean 646.2) and \"Shop\" (mean 10633) is much higher than for \"Private\" user_type (35.3).
", "votes": null, "replies": [ { "id": 326605, "author_name": "cgundlach", "author_url": "", "post_date": "05/10/2018 02:26:38", "content": "That would make sense if that is what it is, and why it could work as either categorical or numerical
", "votes": null, "replies": [] }, { "id": 326607, "author_name": "peterhurford", "author_url": "", "post_date": "05/10/2018 02:28:38", "content": "If my interpretation is correct, it is technically an \"ordinal\". So far I've been treating it as numeric.
", "votes": null, "replies": [] }, { "id": 329991, "author_name": "shanth84", "author_url": "", "post_date": "05/17/2018 19:02:10", "content": "@ Peter \nI am treating this as a numeric myself. IS that how you are doing it for your current LB position?
", "votes": null, "replies": [] } ] }, { "id": 326663, "author_name": "ydhiman", "author_url": "", "post_date": "05/10/2018 05:51:58", "content": "Agree ! with Peter that this variable can be the sequence of images uploaded by the user but an addition to it is that its since registration . Most of the user item_seq_number starts from 1 . Some which has item_sequence_number more than 1 means that , first image loaded by them is in the past and is not present in train set time duration . I believe 72385 users (i.e. ~ 9% ) has a first time post on the website in train set duration.
\n\n(train.groupby('user_id')['item_seq_number'].max() <= 1).value_counts()
\n\nCorrect me if my understanding is wrong .
", "votes": null, "replies": [] }, { "id": 326716, "author_name": "steinhafen", "author_url": "", "post_date": "05/10/2018 08:16:03", "content": "Its something like AD ID system, check this out http://www.ad-id.org/how-it-works/ad-id-structure
", "votes": null, "replies": [] }, { "id": 329992, "author_name": "shanth84", "author_url": "", "post_date": "05/17/2018 19:07:36", "content": "Found another discussion, that is talking about interpreting this variable. \nThe user talks about how this can be, \"A Sequential number in which the specific item_id (unique AD) will be displayed to the same user_id across multiple devices\"
\n\nDiscussion link is here How to interpret item_seq_number
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "326595": "The definition the competition gives us is: \"item_seq_number - Ad sequential number for user.\". That seems very vague to me, is it a numerical or categorical feature?", "326602": "I agree it is vague and clarification on this variable, and many other variables (including the target!) would be helpful.\n\nFrom looking at the variable, my best guess is that an item_seq_number of X means that is the Xth ad the user posted.\n\nI infer this because:\n(a) item_seq_number is higher as time is higher and\n(b) mean item_seq_number is much higher for user_types I expect post more -- \"Company\" (mean 646.2) and \"Shop\" (mean 10633) is much higher than for \"Private\" user_type (35.3).", "326605": "That would make sense if that is what it is, and why it could work as either categorical or numerical", "326607": "If my interpretation is correct, it is technically an \"ordinal\". So far I've been treating it as numeric.", "326663": "Agree ! with Peter that this variable can be the sequence of images uploaded by the user but an addition to it is that its since registration . Most of the user item_seq_number starts from 1 . Some which has item_sequence_number more than 1 means that , first image loaded by them is in the past and is not present in train set time duration . I believe 72385 users (i.e. ~ 9% ) has a first time post on the website in train set duration. \n\n(train.groupby('user_id')['item_seq_number'].max() <= 1).value_counts()\n\nCorrect me if my understanding is wrong .", "326716": "Its something like AD ID system, check this out http://www.ad-id.org/how-it-works/ad-id-structure", "329991": "Peter \nI am treating this as a numeric myself. IS that how you are doing it for your current LB position?", "329992": "Found another discussion, that is talking about interpreting this variable. \nThe user talks about how this can be, \"A Sequential number in which the specific item_id (unique AD) will be displayed to the same user_id across multiple devices\"\n\nDiscussion link is here [How to interpret item_seq_number][1]\n\n\n [1]: https://www.kaggle.com/c/avito-demand-prediction/discussion/56693" }, "source": "meta" }