{ "id": 315054, "title": "Questions about the dataset", "url": "/competitions/h-and-m-personalized-fashion-recommendations/discussion/315054", "author_name": "", "post_date": "2022-03-25T21:41:59.365002200Z", "votes": null, "comment_count": 5, "views": 0, "content": "
    \n

  1. for the sample_submissions, is it the 1st most likely product to the 12th most likely product, or does order not actually matter?

    2. \n

  2. I cannot tell if a unique customer will always have the same customer_id (1 to 1) throughout, or… if a unique customer will have many customer_id's, and if so, whether 2 different people can have the same customer_id's.

    3. \n

  3. When they mean 7-day period, that means starting on the day right after the last day of the training data, right?

    4. \n
\n

Thank you.

", "messages": [ { "id": "1735116", "postDate": "03/25/2022 21:41:59", "content": "
    \n

  1. for the sample_submissions, is it the 1st most likely product to the 12th most likely product, or does order not actually matter?

    2. \n

  2. I cannot tell if a unique customer will always have the same customer_id (1 to 1) throughout, or… if a unique customer will have many customer_id's, and if so, whether 2 different people can have the same customer_id's.

    3. \n

  3. When they mean 7-day period, that means starting on the day right after the last day of the training data, right?

    4. \n
\n

Thank you.

", "rawMarkdown": "1. for the sample_submissions, is it the 1st most likely product to the 12th most likely product, or does order not actually matter?\n\n2. I cannot tell if a unique customer will always have the same customer_id (1 to 1) throughout, or... if a unique customer will have many customer_id's, and if so, whether 2 different people can have the same customer_id's.\n\n3. When they mean 7-day period, that means starting on the day right after the last day of the training data, right?\n\nThank you.", "votes": null }, { "id": "1735516", "postDate": "03/26/2022 11:15:42", "content": "

As far as I get, I can answer question 1 and 3:

\n
    \n

  1. The order does matter, the MAP is sensible to the order and the error will be lower if you match the first product rather than the last product

    2. \n

  2. Yes, exactly: prediction time period is from 23rd Sep2020 to 29th Sep2020 for all customers

    3. \n
\n

On the question #2 I fear that to a unique customer_id may correspond 1/2 or more customers, I mean:
\nthe customer_id is an account, i.e. the identifier of the client that has registered him/herself on the webside or owns the H&M card. However, if someone uses the account of another one (e.g. mother & son, girlfriend &boyfriend …) you will have 2 customers using the same customer_id (i.e. same account on the website, same card on physical shop). This is my thoughts.

", "rawMarkdown": "As far as I get, I can answer question 1 and 3:\n1. The order does matter, the MAP is sensible to the order and the error will be lower if you match the first product rather than the last product\n\n3. Yes, exactly: prediction time period is from 23rd Sep2020 to 29th Sep2020 for all customers\n\nOn the question #2 I fear that to a unique customer_id may correspond 1/2 or more customers, I mean:\nthe customer_id is an account, i.e. the identifier of the client that has registered him/herself on the webside or owns the H&M card. However, if someone uses the account of another one (e.g. mother & son, girlfriend &boyfriend ...) you will have 2 customers using the same customer_id (i.e. same account on the website, same card on physical shop). This is my thoughts.", "votes": null }, { "id": "1735674", "postDate": "03/26/2022 14:05:58", "content": "

thanks. your answer is helpful

", "rawMarkdown": "thanks. your answer is helpful", "votes": null }, { "id": "1736710", "postDate": "03/27/2022 16:29:32", "content": "

But to your point @MattiaCenturelli, where is the test data - I mean we have the transaction data in the csv file \"transactions_train.csv\", ONLY upto Sept 22, 2020. In order to use the prediction time period Sept 23 to Sept 29, 2020 where is that data? Am I missing something?

", "rawMarkdown": "But to your point @MattiaCenturelli, where is the test data - I mean we have the transaction data in the csv file \"***transactions_train.csv***\", ONLY upto **Sept 22, 2020**. In order to use the prediction time period Sept 23 to Sept 29, 2020 where is that data? Am I missing something?", "votes": null }, { "id": "1736866", "postDate": "03/27/2022 20:08:16", "content": "

@pankajashankar I think you should compute it, because as the description explains \"transactions_train.csv\" represents training data set, the test set is not there. I am moving Walk Forward (weekly) in order to see my validations, is someone else doing the same?

", "rawMarkdown": "pankajashankar I think you should compute it, because as the description explains \"transactions_train.csv\" represents training data set, the test set is not there. I am moving Walk Forward (weekly) in order to see my validations, is someone else doing the same?", "votes": null }, { "id": "1736877", "postDate": "03/27/2022 20:36:45", "content": "

The test data is private as we would be able to make perfect predictions if it was public.

\n

We need to predict 12 articles for every customer and these predictions will be scored against the private, real purchases of the customers in the testing week. Most of them probably didn't buy anything, but we don't know which ones.

", "rawMarkdown": "The test data is private as we would be able to make perfect predictions if it was public.\n\nWe need to predict 12 articles for every customer and these predictions will be scored against the private, real purchases of the customers in the testing week. Most of them probably didn't buy anything, but we don't know which ones.", "votes": null } ], "comments": [ { "id": 1735516, "author_name": "centu93", "author_url": "", "post_date": "03/26/2022 11:15:42", "content": "

As far as I get, I can answer question 1 and 3:

\n
    \n

  1. The order does matter, the MAP is sensible to the order and the error will be lower if you match the first product rather than the last product

    2. \n

  2. Yes, exactly: prediction time period is from 23rd Sep2020 to 29th Sep2020 for all customers

    3. \n
\n

On the question #2 I fear that to a unique customer_id may correspond 1/2 or more customers, I mean:
\nthe customer_id is an account, i.e. the identifier of the client that has registered him/herself on the webside or owns the H&M card. However, if someone uses the account of another one (e.g. mother & son, girlfriend &boyfriend …) you will have 2 customers using the same customer_id (i.e. same account on the website, same card on physical shop). This is my thoughts.

", "votes": null, "replies": [ { "id": 1735674, "author_name": "deepkun1995", "author_url": "", "post_date": "03/26/2022 14:05:58", "content": "

thanks. your answer is helpful

", "votes": null, "replies": [] } ] }, { "id": 1736710, "author_name": "pankajashankar", "author_url": "", "post_date": "03/27/2022 16:29:32", "content": "

But to your point @MattiaCenturelli, where is the test data - I mean we have the transaction data in the csv file \"transactions_train.csv\", ONLY upto Sept 22, 2020. In order to use the prediction time period Sept 23 to Sept 29, 2020 where is that data? Am I missing something?

", "votes": null, "replies": [ { "id": 1736877, "author_name": "tbierhance", "author_url": "", "post_date": "03/27/2022 20:36:45", "content": "

The test data is private as we would be able to make perfect predictions if it was public.

\n

We need to predict 12 articles for every customer and these predictions will be scored against the private, real purchases of the customers in the testing week. Most of them probably didn't buy anything, but we don't know which ones.

", "votes": null, "replies": [] } ] }, { "id": 1736866, "author_name": "giuseppemartucci452", "author_url": "", "post_date": "03/27/2022 20:08:16", "content": "

@pankajashankar I think you should compute it, because as the description explains \"transactions_train.csv\" represents training data set, the test set is not there. I am moving Walk Forward (weekly) in order to see my validations, is someone else doing the same?

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1735116": "1. for the sample_submissions, is it the 1st most likely product to the 12th most likely product, or does order not actually matter?\n\n2. I cannot tell if a unique customer will always have the same customer_id (1 to 1) throughout, or... if a unique customer will have many customer_id's, and if so, whether 2 different people can have the same customer_id's.\n\n3. When they mean 7-day period, that means starting on the day right after the last day of the training data, right?\n\nThank you.", "1735516": "As far as I get, I can answer question 1 and 3:\n1. The order does matter, the MAP is sensible to the order and the error will be lower if you match the first product rather than the last product\n\n3. Yes, exactly: prediction time period is from 23rd Sep2020 to 29th Sep2020 for all customers\n\nOn the question #2 I fear that to a unique customer_id may correspond 1/2 or more customers, I mean:\nthe customer_id is an account, i.e. the identifier of the client that has registered him/herself on the webside or owns the H&M card. However, if someone uses the account of another one (e.g. mother & son, girlfriend &boyfriend ...) you will have 2 customers using the same customer_id (i.e. same account on the website, same card on physical shop). This is my thoughts.", "1735674": "thanks. your answer is helpful", "1736710": "But to your point @MattiaCenturelli, where is the test data - I mean we have the transaction data in the csv file \"***transactions_train.csv***\", ONLY upto **Sept 22, 2020**. In order to use the prediction time period Sept 23 to Sept 29, 2020 where is that data? Am I missing something?", "1736866": "pankajashankar I think you should compute it, because as the description explains \"transactions_train.csv\" represents training data set, the test set is not there. I am moving Walk Forward (weekly) in order to see my validations, is someone else doing the same?", "1736877": "The test data is private as we would be able to make perfect predictions if it was public.\n\nWe need to predict 12 articles for every customer and these predictions will be scored against the private, real purchases of the customers in the testing week. Most of them probably didn't buy anything, but we don't know which ones." }, "source": "meta" }