Welcome to the Corporación Favorita Grocery Sales Forecasting Challenge! Both Favorita and Kaggle are excited for to host this competition, which provides a challenging data set of over 125 million rows of daily sales data for over 4,000 products sold at 54 different store locations.
\n\nThere are a number of important notes found in the Data Description section of this contest, so make sure you familiarize yourself with this information, as it will answer a number of question you may have.
\n\nPlease feel free to ask additional questions about the contest in this thread.
\n\nGood luck!
", "messages": [ { "id": "233286", "postDate": "10/19/2017 19:22:32", "content": "Welcome to the Corporación Favorita Grocery Sales Forecasting Challenge! Both Favorita and Kaggle are excited for to host this competition, which provides a challenging data set of over 125 million rows of daily sales data for over 4,000 products sold at 54 different store locations.
\n\nThere are a number of important notes found in the Data Description section of this contest, so make sure you familiarize yourself with this information, as it will answer a number of question you may have.
\n\nPlease feel free to ask additional questions about the contest in this thread.
\n\nGood luck!
", "rawMarkdown": "Welcome to the Corporación Favorita Grocery Sales Forecasting Challenge! Both Favorita and Kaggle are excited for to host this competition, which provides a challenging data set of over 125 million rows of daily sales data for over 4,000 products sold at 54 different store locations.\n\nThere are a number of important notes found in the [Data Description section](https://www.kaggle.com/c/favorita-grocery-sales-forecasting/data) of this contest, so make sure you familiarize yourself with this information, as it will answer a number of question you may have.\n\nPlease feel free to ask additional questions about the contest in this thread.\n\nGood luck!", "votes": null }, { "id": "233402", "postDate": "10/20/2017 05:09:58", "content": "Quick question about the evaluation metrics: can you confirm the weights used in the formula are based on the per-item perishable indicator from the items.csv file?
", "rawMarkdown": "Quick question about the evaluation metrics: can you confirm the weights used in the formula are based on the per-item perishable indicator from the items.csv file?", "votes": null }, { "id": "233441", "postDate": "10/20/2017 07:32:01", "content": "what is a store type (A,B,C ... ) ?
", "rawMarkdown": "what is a store type (A,B,C ... ) ?", "votes": null }, { "id": "233536", "postDate": "10/20/2017 13:33:04", "content": "That is correct. Is there a change to the wording that you can suggest to make it more clear? (Or, perhaps, just use what you stated above?)
", "rawMarkdown": "That is correct. Is there a change to the wording that you can suggest to make it more clear? (Or, perhaps, just use what you stated above?)", "votes": null }, { "id": "233690", "postDate": "10/20/2017 20:49:27", "content": "Is not listing price for each item an oversight, or a deliberate omission?
", "rawMarkdown": "Is not listing price for each item an oversight, or a deliberate omission?", "votes": null }, { "id": "233761", "postDate": "10/21/2017 05:02:48", "content": "I don't think there's anything unclear about the wording on the data page, but explicitly defining the weights on the evaluation page would be useful.
", "rawMarkdown": "I don't think there's anything unclear about the wording on the data page, but explicitly defining the weights on the evaluation page would be useful.", "votes": null }, { "id": "233984", "postDate": "10/22/2017 01:06:33", "content": "so basically, weight = 1.25 for perishable items and weight = 1 for non-perishable items, as mentioned in data description for items.csv
", "rawMarkdown": "so basically, weight = 1.25 for perishable items and weight = 1 for non-perishable items, as mentioned in data description for items.csv", "votes": null }, { "id": "234371", "postDate": "10/23/2017 06:17:30", "content": "Why is the test data only the last two weeks, when there was so much data to work with? Interesting/odd to me, that.
", "rawMarkdown": "Why is the test data only the last two weeks, when there was so much data to work with? Interesting/odd to me, that.", "votes": null }, { "id": "234510", "postDate": "10/23/2017 14:08:56", "content": "Same question wandering in my mind. I asked this on a similar discussion on this forum. Nearly 97% of the data is the training set, which is of course odd since it seems to be violating the 70:30 rule.
", "rawMarkdown": "Same question wandering in my mind. I asked this on a similar discussion on this forum. Nearly 97% of the data is the training set, which is of course odd since it seems to be violating the 70:30 rule.", "votes": null }, { "id": "234527", "postDate": "10/23/2017 14:48:59", "content": "Please clarify which day or days the unit sales submission is supposed to cover. Thank you
", "rawMarkdown": "Please clarify which day or days the unit sales submission is supposed to cover. Thank you", "votes": null }, { "id": "234557", "postDate": "10/23/2017 15:25:02", "content": "OK, so it was not clear, but it seems that training covers 2013-2016 and that the test data is a 2 week window in 2017. You would get much better outputs from continuous time series data, and the inclusion of price data.
", "rawMarkdown": "OK, so it was not clear, but it seems that training covers 2013-2016 and that the test data is a 2 week window in 2017. You would get much better outputs from continuous time series data, and the inclusion of price data.", "votes": null }, { "id": "234663", "postDate": "10/23/2017 19:40:41", "content": "I've updated the evaluation page to include the statement about item weights, so it is more clear where they are coming from.
", "rawMarkdown": "I've updated the [evaluation page](https://www.kaggle.com/c/favorita-grocery-sales-forecasting#evaluation) to include the statement about item weights, so it is more clear where they are coming from.", "votes": null }, { "id": "235218", "postDate": "10/24/2017 23:44:52", "content": "Quick question. So it's easy to spot 'new-market-trial' items in the test dataset that is nowhere in the training dataset. My question is, for those items that DO present in the training set but not in a particular store (seems like not all stores are stocked with the same number of unique items), should we assume that those items do not exist in such store, thus our prediction for the unit_sales of that store/item combination is 0. Or should we also treat them as 'new-market-trial' items, and predict their stock based on similar items?
", "rawMarkdown": "Quick question. So it's easy to spot 'new-market-trial' items in the test dataset that is nowhere in the training dataset. My question is, for those items that DO present in the training set but not in a particular store (seems like not all stores are stocked with the same number of unique items), should we assume that those items do not exist in such store, thus our prediction for the unit_sales of that store/item combination is 0. Or should we also treat them as 'new-market-trial' items, and predict their stock based on similar items?", "votes": null }, { "id": "235291", "postDate": "10/25/2017 03:55:44", "content": "We use 16 days for the test set because that is our real life business case. We don´t need to predict further in the future.
", "rawMarkdown": "We use 16 days for the test set because that is our real life business case. We don´t need to predict further in the future.", "votes": null }, { "id": "235293", "postDate": "10/25/2017 04:06:45", "content": "The store types are the different store banners under which our supermarkets operate.
", "rawMarkdown": "The store types are the different store banners under which our supermarkets operate.", "votes": null }, { "id": "235294", "postDate": "10/25/2017 04:08:55", "content": "It is a deliberate omission because that is highly sensible information that cannot be used in the competition.
", "rawMarkdown": "It is a deliberate omission because that is highly sensible information that cannot be used in the competition.", "votes": null }, { "id": "530480", "postDate": "05/13/2019 01:35:47", "content": "hello i want to ask a privacy question , is it allowed to use the data as an example to writing books about meachine learning?
", "rawMarkdown": "hello i want to ask a privacy question , is it allowed to use the data as an example to writing books about meachine learning?", "votes": null }, { "id": "1556264", "postDate": "10/24/2021 18:54:23", "content": "Hi! Please, anyone, Tell me! what is a store type (A,B,C … ) ?
", "rawMarkdown": "Hi! Please, anyone, Tell me! what is a store type (A,B,C … ) ?", "votes": null } ], "comments": [ { "id": 1556264, "author_name": "jillanisofttech", "author_url": "", "post_date": "10/24/2021 18:54:23", "content": "Hi! Please, anyone, Tell me! what is a store type (A,B,C … ) ?
", "votes": null, "replies": [] }, { "id": 233402, "author_name": "konradb", "author_url": "", "post_date": "10/20/2017 05:09:58", "content": "Quick question about the evaluation metrics: can you confirm the weights used in the formula are based on the per-item perishable indicator from the items.csv file?
", "votes": null, "replies": [ { "id": 233536, "author_name": "inversion", "author_url": "", "post_date": "10/20/2017 13:33:04", "content": "That is correct. Is there a change to the wording that you can suggest to make it more clear? (Or, perhaps, just use what you stated above?)
", "votes": null, "replies": [] }, { "id": 233761, "author_name": "joconnor", "author_url": "", "post_date": "10/21/2017 05:02:48", "content": "I don't think there's anything unclear about the wording on the data page, but explicitly defining the weights on the evaluation page would be useful.
", "votes": null, "replies": [] }, { "id": 233984, "author_name": "madankan", "author_url": "", "post_date": "10/22/2017 01:06:33", "content": "so basically, weight = 1.25 for perishable items and weight = 1 for non-perishable items, as mentioned in data description for items.csv
", "votes": null, "replies": [] }, { "id": 234663, "author_name": "inversion", "author_url": "", "post_date": "10/23/2017 19:40:41", "content": "I've updated the evaluation page to include the statement about item weights, so it is more clear where they are coming from.
", "votes": null, "replies": [] } ] }, { "id": 233441, "author_name": "noordk", "author_url": "", "post_date": "10/20/2017 07:32:01", "content": "what is a store type (A,B,C ... ) ?
", "votes": null, "replies": [ { "id": 235293, "author_name": "gyorgy", "author_url": "", "post_date": "10/25/2017 04:06:45", "content": "The store types are the different store banners under which our supermarkets operate.
", "votes": null, "replies": [] } ] }, { "id": 233690, "author_name": "mathdr", "author_url": "", "post_date": "10/20/2017 20:49:27", "content": "Is not listing price for each item an oversight, or a deliberate omission?
", "votes": null, "replies": [ { "id": 235294, "author_name": "gyorgy", "author_url": "", "post_date": "10/25/2017 04:08:55", "content": "It is a deliberate omission because that is highly sensible information that cannot be used in the competition.
", "votes": null, "replies": [] } ] }, { "id": 234371, "author_name": "happycube", "author_url": "", "post_date": "10/23/2017 06:17:30", "content": "Why is the test data only the last two weeks, when there was so much data to work with? Interesting/odd to me, that.
", "votes": null, "replies": [ { "id": 234510, "author_name": "shivendra90", "author_url": "", "post_date": "10/23/2017 14:08:56", "content": "Same question wandering in my mind. I asked this on a similar discussion on this forum. Nearly 97% of the data is the training set, which is of course odd since it seems to be violating the 70:30 rule.
", "votes": null, "replies": [] }, { "id": 235291, "author_name": "gyorgy", "author_url": "", "post_date": "10/25/2017 03:55:44", "content": "We use 16 days for the test set because that is our real life business case. We don´t need to predict further in the future.
", "votes": null, "replies": [] } ] }, { "id": 234527, "author_name": "wilkinsondi", "author_url": "", "post_date": "10/23/2017 14:48:59", "content": "Please clarify which day or days the unit sales submission is supposed to cover. Thank you
", "votes": null, "replies": [ { "id": 234557, "author_name": "wilkinsondi", "author_url": "", "post_date": "10/23/2017 15:25:02", "content": "OK, so it was not clear, but it seems that training covers 2013-2016 and that the test data is a 2 week window in 2017. You would get much better outputs from continuous time series data, and the inclusion of price data.
", "votes": null, "replies": [] } ] }, { "id": 235218, "author_name": "sageknight", "author_url": "", "post_date": "10/24/2017 23:44:52", "content": "Quick question. So it's easy to spot 'new-market-trial' items in the test dataset that is nowhere in the training dataset. My question is, for those items that DO present in the training set but not in a particular store (seems like not all stores are stocked with the same number of unique items), should we assume that those items do not exist in such store, thus our prediction for the unit_sales of that store/item combination is 0. Or should we also treat them as 'new-market-trial' items, and predict their stock based on similar items?
", "votes": null, "replies": [] }, { "id": 530480, "author_name": "caijiangyao1991", "author_url": "", "post_date": "05/13/2019 01:35:47", "content": "hello i want to ask a privacy question , is it allowed to use the data as an example to writing books about meachine learning?
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "233286": "Welcome to the Corporación Favorita Grocery Sales Forecasting Challenge! Both Favorita and Kaggle are excited for to host this competition, which provides a challenging data set of over 125 million rows of daily sales data for over 4,000 products sold at 54 different store locations.\n\nThere are a number of important notes found in the [Data Description section](https://www.kaggle.com/c/favorita-grocery-sales-forecasting/data) of this contest, so make sure you familiarize yourself with this information, as it will answer a number of question you may have.\n\nPlease feel free to ask additional questions about the contest in this thread.\n\nGood luck!", "233402": "Quick question about the evaluation metrics: can you confirm the weights used in the formula are based on the per-item perishable indicator from the items.csv file?", "233441": "what is a store type (A,B,C ... ) ?", "233536": "That is correct. Is there a change to the wording that you can suggest to make it more clear? (Or, perhaps, just use what you stated above?)", "233690": "Is not listing price for each item an oversight, or a deliberate omission?", "233761": "I don't think there's anything unclear about the wording on the data page, but explicitly defining the weights on the evaluation page would be useful.", "233984": "so basically, weight = 1.25 for perishable items and weight = 1 for non-perishable items, as mentioned in data description for items.csv", "234371": "Why is the test data only the last two weeks, when there was so much data to work with? Interesting/odd to me, that.", "234510": "Same question wandering in my mind. I asked this on a similar discussion on this forum. Nearly 97% of the data is the training set, which is of course odd since it seems to be violating the 70:30 rule.", "234527": "Please clarify which day or days the unit sales submission is supposed to cover. Thank you", "234557": "OK, so it was not clear, but it seems that training covers 2013-2016 and that the test data is a 2 week window in 2017. You would get much better outputs from continuous time series data, and the inclusion of price data.", "234663": "I've updated the [evaluation page](https://www.kaggle.com/c/favorita-grocery-sales-forecasting#evaluation) to include the statement about item weights, so it is more clear where they are coming from.", "235218": "Quick question. So it's easy to spot 'new-market-trial' items in the test dataset that is nowhere in the training dataset. My question is, for those items that DO present in the training set but not in a particular store (seems like not all stores are stocked with the same number of unique items), should we assume that those items do not exist in such store, thus our prediction for the unit_sales of that store/item combination is 0. Or should we also treat them as 'new-market-trial' items, and predict their stock based on similar items?", "235291": "We use 16 days for the test set because that is our real life business case. We don´t need to predict further in the future.", "235293": "The store types are the different store banners under which our supermarkets operate.", "235294": "It is a deliberate omission because that is highly sensible information that cannot be used in the competition.", "530480": "hello i want to ask a privacy question , is it allowed to use the data as an example to writing books about meachine learning?", "1556264": "Hi! Please, anyone, Tell me! what is a store type (A,B,C … ) ?" }, "source": "meta" }