Hello,\nI see that variables is_booking and cnt are missing in the test set. Why is it so? Are we supposed to treat them as target values too?\nThanks.
", "messages": [ { "id": "116083", "postDate": "04/21/2016 22:53:01", "content": "Hello,\nI see that variables is_booking and cnt are missing in the test set. Why is it so? Are we supposed to treat them as target values too?\nThanks.
", "rawMarkdown": "Hello,\r\nI see that variables is_booking and cnt are missing in the test set. Why is it so? Are we supposed to treat them as target values too?\r\nThanks.", "votes": null }, { "id": "116087", "postDate": "04/21/2016 23:02:58", "content": "You can find in the data description info about it:
\n\n\n", "rawMarkdown": "You can find in the data description info about it:\r\n> Test data only includes booking events.", "votes": null }, { "id": "116101", "postDate": "04/22/2016 02:01:22", "content": "Test data only includes booking events.
\n
Actually, I was not sure why the variables is_booking exist in the train data.
\n\nIf we should submit 5 hotel_cluster for each id, it makes no difference whether is_booking is 1 or 0.
\n\nI mean this competition is cluster prediction, but not booking prediction.
", "rawMarkdown": "Actually, I was not sure why the variables is_booking exist in the train data.\r\n\r\nIf we should submit 5 hotel_cluster for each id, it makes no difference whether is_booking is 1 or 0.\r\n\r\nI mean this competition is cluster prediction, but not booking prediction.", "votes": null }, { "id": "116125", "postDate": "04/22/2016 06:14:12", "content": "[quote=ONODERA;116101]\nActually, I was not sure why the variables is_booking exist in the train data.\n[/quote]
\n\nI think is_booking is an important variable.
\n\nI imagine that when building a model the rows with is_booking == 1 should be weighted to give the optimal result, since ultimately we are predicting hotel cluster only for bookings.
\n\nAlso, I think whatever the model it might be a good idea to tune using only is_booking == 1 rows in the validation set.
\n\nSo as I see it the rows with is_booking == 0 (clicks) are just something extra for the training which might help (most likely they will) or which might not. But how to use them is what has to be investigated.
", "rawMarkdown": "[quote=ONODERA;116101]\r\nActually, I was not sure why the variables is_booking exist in the train data.\r\n[/quote]\r\n\r\nI think is_booking is an important variable. \r\n\r\nI imagine that when building a model the rows with is_booking == 1 should be weighted to give the optimal result, since ultimately we are predicting hotel cluster only for bookings.\r\n\r\nAlso, I think whatever the model it might be a good idea to tune using only is_booking == 1 rows in the validation set. \r\n\r\nSo as I see it the rows with is_booking == 0 (clicks) are just something extra for the training which might help (most likely they will) or which might not. But how to use them is what has to be investigated.", "votes": null }, { "id": "116562", "postDate": "04/25/2016 06:35:21", "content": "is_booking is meaningful as users' booking behavior is much more important than their click behaviors. However, what does 'cnt' means?
", "rawMarkdown": "is_booking is meaningful as users' booking behavior is much more important than their click behaviors. However, what does 'cnt' means?", "votes": null }, { "id": "116581", "postDate": "04/25/2016 09:14:48", "content": "If a user A clicks a hotel B 5 times is the same session then cnt for the corresponding record will be 5.
\n\nKeep in mind that there can be different types of clicks (e.g. clicks on pictures, user reviews, room rates, etc.); feel free to give it a try on the expedia.com site. However, for the purpose of this of this competition we don't distinguish different types of clicks.
", "rawMarkdown": "If a user A clicks a hotel B 5 times is the same session then cnt for the corresponding record will be 5.\r\n\r\nKeep in mind that there can be different types of clicks (e.g. clicks on pictures, user reviews, room rates, etc.); feel free to give it a try on the expedia.com site. However, for the purpose of this of this competition we don't distinguish different types of clicks.", "votes": null }, { "id": "120317", "postDate": "05/17/2016 11:44:30", "content": "What is the answer for the variables which are missing? Were they supposed to be present in the test and training data?
", "rawMarkdown": "What is the answer for the variables which are missing? Were they supposed to be present in the test and training data?", "votes": null } ], "comments": [ { "id": 116087, "author_name": "valeriyischenko", "author_url": "", "post_date": "04/21/2016 23:02:58", "content": "You can find in the data description info about it:
\n\n\n", "votes": null, "replies": [] }, { "id": 116101, "author_name": "onodera", "author_url": "", "post_date": "04/22/2016 02:01:22", "content": "Test data only includes booking events.
\n
Actually, I was not sure why the variables is_booking exist in the train data.
\n\nIf we should submit 5 hotel_cluster for each id, it makes no difference whether is_booking is 1 or 0.
\n\nI mean this competition is cluster prediction, but not booking prediction.
", "votes": null, "replies": [] }, { "id": 116125, "author_name": "pekkaa", "author_url": "", "post_date": "04/22/2016 06:14:12", "content": "[quote=ONODERA;116101]\nActually, I was not sure why the variables is_booking exist in the train data.\n[/quote]
\n\nI think is_booking is an important variable.
\n\nI imagine that when building a model the rows with is_booking == 1 should be weighted to give the optimal result, since ultimately we are predicting hotel cluster only for bookings.
\n\nAlso, I think whatever the model it might be a good idea to tune using only is_booking == 1 rows in the validation set.
\n\nSo as I see it the rows with is_booking == 0 (clicks) are just something extra for the training which might help (most likely they will) or which might not. But how to use them is what has to be investigated.
", "votes": null, "replies": [] }, { "id": 116562, "author_name": "ifleezju", "author_url": "", "post_date": "04/25/2016 06:35:21", "content": "is_booking is meaningful as users' booking behavior is much more important than their click behaviors. However, what does 'cnt' means?
", "votes": null, "replies": [] }, { "id": 116581, "author_name": "adamwoz", "author_url": "", "post_date": "04/25/2016 09:14:48", "content": "If a user A clicks a hotel B 5 times is the same session then cnt for the corresponding record will be 5.
\n\nKeep in mind that there can be different types of clicks (e.g. clicks on pictures, user reviews, room rates, etc.); feel free to give it a try on the expedia.com site. However, for the purpose of this of this competition we don't distinguish different types of clicks.
", "votes": null, "replies": [] }, { "id": 120317, "author_name": "vinaylad", "author_url": "", "post_date": "05/17/2016 11:44:30", "content": "What is the answer for the variables which are missing? Were they supposed to be present in the test and training data?
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "116083": "Hello,\r\nI see that variables is_booking and cnt are missing in the test set. Why is it so? Are we supposed to treat them as target values too?\r\nThanks.", "116087": "You can find in the data description info about it:\r\n> Test data only includes booking events.", "116101": "Actually, I was not sure why the variables is_booking exist in the train data.\r\n\r\nIf we should submit 5 hotel_cluster for each id, it makes no difference whether is_booking is 1 or 0.\r\n\r\nI mean this competition is cluster prediction, but not booking prediction.", "116125": "[quote=ONODERA;116101]\r\nActually, I was not sure why the variables is_booking exist in the train data.\r\n[/quote]\r\n\r\nI think is_booking is an important variable. \r\n\r\nI imagine that when building a model the rows with is_booking == 1 should be weighted to give the optimal result, since ultimately we are predicting hotel cluster only for bookings.\r\n\r\nAlso, I think whatever the model it might be a good idea to tune using only is_booking == 1 rows in the validation set. \r\n\r\nSo as I see it the rows with is_booking == 0 (clicks) are just something extra for the training which might help (most likely they will) or which might not. But how to use them is what has to be investigated.", "116562": "is_booking is meaningful as users' booking behavior is much more important than their click behaviors. However, what does 'cnt' means?", "116581": "If a user A clicks a hotel B 5 times is the same session then cnt for the corresponding record will be 5.\r\n\r\nKeep in mind that there can be different types of clicks (e.g. clicks on pictures, user reviews, room rates, etc.); feel free to give it a try on the expedia.com site. However, for the purpose of this of this competition we don't distinguish different types of clicks.", "120317": "What is the answer for the variables which are missing? Were they supposed to be present in the test and training data?" }, "source": "meta" }