{ "id": 315587, "title": "Validation for @astrung insight", "url": "/competitions/h-and-m-personalized-fashion-recommendations/discussion/315587", "author_name": "Y-Haneji", "post_date": "2022-03-29T00:54:24.076000", "votes": 6, "comment_count": 0, "views": 0, "content": "

@astrung did great research about the user's distribution.

\n

Discussion: https://www.kaggle.com/competitions/h-and-m-personalized-fashion-recommendations/discussion/312653
\nNotebook: https://www.kaggle.com/astrung/eda-extract-user-metadata-to-apply-deep-model/notebook
\nPlease UPVOTE them too!
\nThanks for your great interesting Notebooks and Notebooks!

\n

I validated its strategy by hold-out method.
\nValid term is from 2020-09-16 to 2020-09-22 (both included).

\n

The status is labeled by using transactions before valid term. (to avoid leak)
\nThe percent focuses on only users who bought any item in valid term.

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
columnvaluepercent
lastest_inactive_months039.89
120.31
211.15
328.65
---------
active statusactive39.89
inactive_in_1_month20.31
inactive_in_2_months11.15
inactive_in_3_months_or_more14.95
inactive_in_year13.71
---------
cold_start_statuscold_start40.11
non_cold_start59.89
\n

active & non cold start user (by all users who bought any item in valid term): 32.97%

\n

The active & non cold start user's percentage is only 9% if calculated by all users but the percentage increases dramatically if calculated by users who bought any item in valid term.

\n
\n

Deep model only works with active/non cold start users. But in our test data, there are only 9% users who satisfy this condition. So can we give up deep model ?

\n
\n

I can say that deep model has great potentiality to improve the score.

\n

Notebook: https://www.kaggle.com/hanejiyuto/validate-for-astrung-insight

", "messages": [ { "id": 1738017, "postDate": "2022-03-29T00:54:24.077Z", "content": "

@astrung did great research about the user's distribution.

\n

Discussion: https://www.kaggle.com/competitions/h-and-m-personalized-fashion-recommendations/discussion/312653
\nNotebook: https://www.kaggle.com/astrung/eda-extract-user-metadata-to-apply-deep-model/notebook
\nPlease UPVOTE them too!
\nThanks for your great interesting Notebooks and Notebooks!

\n

I validated its strategy by hold-out method.
\nValid term is from 2020-09-16 to 2020-09-22 (both included).

\n

The status is labeled by using transactions before valid term. (to avoid leak)
\nThe percent focuses on only users who bought any item in valid term.

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
columnvaluepercent
lastest_inactive_months039.89
120.31
211.15
328.65
---------
active statusactive39.89
inactive_in_1_month20.31
inactive_in_2_months11.15
inactive_in_3_months_or_more14.95
inactive_in_year13.71
---------
cold_start_statuscold_start40.11
non_cold_start59.89
\n

active & non cold start user (by all users who bought any item in valid term): 32.97%

\n

The active & non cold start user's percentage is only 9% if calculated by all users but the percentage increases dramatically if calculated by users who bought any item in valid term.

\n
\n

Deep model only works with active/non cold start users. But in our test data, there are only 9% users who satisfy this condition. So can we give up deep model ?

\n
\n

I can say that deep model has great potentiality to improve the score.

\n

Notebook: https://www.kaggle.com/hanejiyuto/validate-for-astrung-insight

", "rawMarkdown": "@astrung did great research about the user's distribution.\n\nDiscussion: https://www.kaggle.com/competitions/h-and-m-personalized-fashion-recommendations/discussion/312653\nNotebook: https://www.kaggle.com/astrung/eda-extract-user-metadata-to-apply-deep-model/notebook\nPlease **UPVOTE** them too!\nThanks for your great interesting Notebooks and Notebooks!\n\nI validated its strategy by hold-out method.\nValid term is from 2020-09-16 to 2020-09-22 (both included).\n\nThe status is labeled by **using transactions before valid term**. (to avoid leak)\nThe percent focuses on **only users who bought any item in valid term**.\n\n| column | value | percent |\n| --- | --- | --- |\n| lastest_inactive_months | 0 | 39.89 |\n| | 1\t| 20.31 |\n| | 2 | 11.15 |\n| | 3 | 28.65 |\n| --- | --- | --- |\n| active status | active | 39.89 |\n| | inactive_in_1_month | 20.31 |\n| | inactive_in_2_months | 11.15 |\n| | inactive_in_3_months_or_more | 14.95 |\n| | inactive_in_year | 13.71 |\n| --- | --- | --- |\n| cold_start_status | cold_start | 40.11 |\n| | non_cold_start | 59.89 |\n\n`active & non cold start user (by all users who bought any item in valid term): 32.97%`\n\nThe active & non cold start user's percentage is **only 9% if calculated by all users** but the percentage **increases dramatically if calculated by users who bought any item in valid term**.\n\n> Deep model only works with active/non cold start users. But in our test data, there are only 9% users who satisfy this condition. So can we give up deep model ?\n\nI can say that **deep model has great potentiality** to improve the score. \n\nNotebook: https://www.kaggle.com/hanejiyuto/validate-for-astrung-insight", "votes": 6 } ], "comments": [], "raw_markdown_by_id": { "1738017": "@astrung did great research about the user's distribution.\n\nDiscussion: https://www.kaggle.com/competitions/h-and-m-personalized-fashion-recommendations/discussion/312653\nNotebook: https://www.kaggle.com/astrung/eda-extract-user-metadata-to-apply-deep-model/notebook\nPlease **UPVOTE** them too!\nThanks for your great interesting Notebooks and Notebooks!\n\nI validated its strategy by hold-out method.\nValid term is from 2020-09-16 to 2020-09-22 (both included).\n\nThe status is labeled by **using transactions before valid term**. (to avoid leak)\nThe percent focuses on **only users who bought any item in valid term**.\n\n| column | value | percent |\n| --- | --- | --- |\n| lastest_inactive_months | 0 | 39.89 |\n| | 1\t| 20.31 |\n| | 2 | 11.15 |\n| | 3 | 28.65 |\n| --- | --- | --- |\n| active status | active | 39.89 |\n| | inactive_in_1_month | 20.31 |\n| | inactive_in_2_months | 11.15 |\n| | inactive_in_3_months_or_more | 14.95 |\n| | inactive_in_year | 13.71 |\n| --- | --- | --- |\n| cold_start_status | cold_start | 40.11 |\n| | non_cold_start | 59.89 |\n\n`active & non cold start user (by all users who bought any item in valid term): 32.97%`\n\nThe active & non cold start user's percentage is **only 9% if calculated by all users** but the percentage **increases dramatically if calculated by users who bought any item in valid term**.\n\n> Deep model only works with active/non cold start users. But in our test data, there are only 9% users who satisfy this condition. So can we give up deep model ?\n\nI can say that **deep model has great potentiality** to improve the score. \n\nNotebook: https://www.kaggle.com/hanejiyuto/validate-for-astrung-insight" } }