Hi \nI am starting with this competition and trying to understand the data. I like to understand things from real world point of view. Below are my doubts along with my understanding (incorrect may be), so please correct me if i am wrong:\nWhat is a card_id ? Is it a person who purchased that card ?
\n\nWhat is a merchant_id ? Is it like, if I used the above card to pay for uber ride, so uber is a merchant ?
\n\nHistorical transaction: this data contains transaction information made by card_id for available merchant_id ?
\n\nThen, who are new_merchants ?
\n\nWhat is the difference between merchants and new_merchants table?
\n\nWhat is first_active_month feature in train data ? Is it the very first month when a person received the card ?
Hi \nI am starting with this competition and trying to understand the data. I like to understand things from real world point of view. Below are my doubts along with my understanding (incorrect may be), so please correct me if i am wrong:\nWhat is a card_id ? Is it a person who purchased that card ?
\n\nWhat is a merchant_id ? Is it like, if I used the above card to pay for uber ride, so uber is a merchant ?
\n\nHistorical transaction: this data contains transaction information made by card_id for available merchant_id ?
\n\nThen, who are new_merchants ?
\n\nWhat is the difference between merchants and new_merchants table?
\n\nWhat is first_active_month feature in train data ? Is it the very first month when a person received the card ?
You are correct about card id, merchant id and Historical transaction.\nMerchant file contains information about the merchants. New merchant file has information about the cards transaction at new merchant (New merchant are those where the card has not visited over a period of 2 months.\nFirst active month is the month when the first purchase was done by the customer (card holder).\nCheck this for the details. There is one data dictionary which can help you a bit with few features as majority of features are anonymous.
\n\nhttps://www.kaggle.com/c/elo-merchant-category-recommendation/data
", "rawMarkdown": "You are correct about card id, merchant id and Historical transaction.\nMerchant file contains information about the merchants. New merchant file has information about the cards transaction at new merchant (New merchant are those where the card has not visited over a period of 2 months.\nFirst active month is the month when the first purchase was done by the customer (card holder).\nCheck this for the details. There is one data dictionary which can help you a bit with few features as majority of features are anonymous.\n\nhttps://www.kaggle.com/c/elo-merchant-category-recommendation/data", "votes": null }, { "id": "443287", "postDate": "12/21/2018 11:04:46", "content": "Hi Harshit,
\n\n Good explantion, but if new merchant are those where the card has not visited over a period of 2 months, How explain this? Merchant, 'M_ID_b029faeccc', was found in both files where the count o month is less then 2? There are others examples, too.\nThanks, if the card has not been visited, how could new merchant information be used to predict loyalty ?
", "rawMarkdown": "Thanks, if the card has not been visited, how could new merchant information be used to predict loyalty ?", "votes": null }, { "id": "443461", "postDate": "12/21/2018 17:29:26", "content": "Fabio, I agree with you. I also see such cases. I am super confused now.
", "rawMarkdown": "Fabio, I agree with you. I also see such cases. I am super confused now.", "votes": null }, { "id": "443952", "postDate": "12/22/2018 19:23:26", "content": "Hi Fabio, I agree on that, may be we can say that those instances are not correct or that is bad data.
\n\nErlich, \nYou can do that, by mapping the features of new transaction with historical transactions.\nConsider a recommendation problem, \nsuppose you visited restaurant A and B and rated them based on 3 factors, i.e \"Ambiance\", \"food quality\", \"Live Music\". Suppose you have high preference for \"Live Music\", therefore gave high overall rating to A( assume it had live music and B didn't). Now, there is another restaurant C (has no live music ),for which someone wants to predict your rating, based on your restaurant A and B analysis, when can say you won't show great interest in going there.\nYou can relate that thing over here.\nHope that helps!
", "rawMarkdown": "Hi Fabio, I agree on that, may be we can say that those instances are not correct or that is bad data. \n\nErlich, \nYou can do that, by mapping the features of new transaction with historical transactions.\nConsider a recommendation problem, \nsuppose you visited restaurant A and B and rated them based on 3 factors, i.e \"Ambiance\", \"food quality\", \"Live Music\". Suppose you have high preference for \"Live Music\", therefore gave high overall rating to A( assume it had live music and B didn't). Now, there is another restaurant C (has no live music ),for which someone wants to predict your rating, based on your restaurant A and B analysis, when can say you won't show great interest in going there.\nYou can relate that thing over here.\nHope that helps!", "votes": null } ], "comments": [ { "id": 442852, "author_name": "harshit92", "author_url": "", "post_date": "12/20/2018 15:59:26", "content": "You are correct about card id, merchant id and Historical transaction.\nMerchant file contains information about the merchants. New merchant file has information about the cards transaction at new merchant (New merchant are those where the card has not visited over a period of 2 months.\nFirst active month is the month when the first purchase was done by the customer (card holder).\nCheck this for the details. There is one data dictionary which can help you a bit with few features as majority of features are anonymous.
\n\nhttps://www.kaggle.com/c/elo-merchant-category-recommendation/data
", "votes": null, "replies": [ { "id": 443287, "author_name": "randrak", "author_url": "", "post_date": "12/21/2018 11:04:46", "content": "Hi Harshit,
\n\n Good explantion, but if new merchant are those where the card has not visited over a period of 2 months, How explain this? Merchant, 'M_ID_b029faeccc', was found in both files where the count o month is less then 2? There are others examples, too.\nThanks, if the card has not been visited, how could new merchant information be used to predict loyalty ?
", "votes": null, "replies": [] }, { "id": 443461, "author_name": "manishsaraswat", "author_url": "", "post_date": "12/21/2018 17:29:26", "content": "Fabio, I agree with you. I also see such cases. I am super confused now.
", "votes": null, "replies": [] }, { "id": 443952, "author_name": "harshit92", "author_url": "", "post_date": "12/22/2018 19:23:26", "content": "Hi Fabio, I agree on that, may be we can say that those instances are not correct or that is bad data.
\n\nErlich, \nYou can do that, by mapping the features of new transaction with historical transactions.\nConsider a recommendation problem, \nsuppose you visited restaurant A and B and rated them based on 3 factors, i.e \"Ambiance\", \"food quality\", \"Live Music\". Suppose you have high preference for \"Live Music\", therefore gave high overall rating to A( assume it had live music and B didn't). Now, there is another restaurant C (has no live music ),for which someone wants to predict your rating, based on your restaurant A and B analysis, when can say you won't show great interest in going there.\nYou can relate that thing over here.\nHope that helps!
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "442765": "Hi \nI am starting with this competition and trying to understand the data. I like to understand things from real world point of view. Below are my doubts along with my understanding (incorrect may be), so please correct me if i am wrong:\nWhat is a card_id ? Is it a person who purchased that card ? \n\nWhat is a merchant_id ? Is it like, if I used the above card to pay for uber ride, so uber is a merchant ?\n\nHistorical transaction: this data contains transaction information made by card_id for available merchant_id ?\n\nThen, who are new_merchants ? \n\nWhat is the difference between merchants and new_merchants table? \n\nWhat is `first_active_month` feature in train data ? Is it the very first month when a person received the card ?", "442852": "You are correct about card id, merchant id and Historical transaction.\nMerchant file contains information about the merchants. New merchant file has information about the cards transaction at new merchant (New merchant are those where the card has not visited over a period of 2 months.\nFirst active month is the month when the first purchase was done by the customer (card holder).\nCheck this for the details. There is one data dictionary which can help you a bit with few features as majority of features are anonymous.\n\nhttps://www.kaggle.com/c/elo-merchant-category-recommendation/data", "443287": "Hi Harshit, \n\n Good explantion, but if new merchant are those where the card has not visited over a period of 2 months, How explain this? Merchant, 'M_ID_b029faeccc', was found in both files where the count o month is less then 2? There are others examples, too.", "443453": "Thanks, if the card has not been visited, how could new merchant information be used to predict loyalty ?", "443461": "Fabio, I agree with you. I also see such cases. I am super confused now.", "443952": "Hi Fabio, I agree on that, may be we can say that those instances are not correct or that is bad data. \n\nErlich, \nYou can do that, by mapping the features of new transaction with historical transactions.\nConsider a recommendation problem, \nsuppose you visited restaurant A and B and rated them based on 3 factors, i.e \"Ambiance\", \"food quality\", \"Live Music\". Suppose you have high preference for \"Live Music\", therefore gave high overall rating to A( assume it had live music and B didn't). Now, there is another restaurant C (has no live music ),for which someone wants to predict your rating, based on your restaurant A and B analysis, when can say you won't show great interest in going there.\nYou can relate that thing over here.\nHope that helps!" }, "source": "meta" }