Welcome to the Expedia Hotel Recommendations competition. The goal is to predict which class of hotel was purchased given past purchases and clicks. This \nis the second time we partner with Kaggle to solve interesting problems (the first competition was https://www.kaggle.com/c/expedia-personalized-sort). \nDon't hesitate to ask any question on the data or the platform. Thanks for participating!
\n\nAdam
", "messages": [ { "id": "115037", "postDate": "04/15/2016 21:03:07", "content": "Welcome to the Expedia Hotel Recommendations competition. The goal is to predict which class of hotel was purchased given past purchases and clicks. This \nis the second time we partner with Kaggle to solve interesting problems (the first competition was https://www.kaggle.com/c/expedia-personalized-sort). \nDon't hesitate to ask any question on the data or the platform. Thanks for participating!
\n\nAdam
", "rawMarkdown": "Welcome to the Expedia Hotel Recommendations competition. The goal is to predict which class of hotel was purchased given past purchases and clicks. This \r\nis the second time we partner with Kaggle to solve interesting problems (the first competition was https://www.kaggle.com/c/expedia-personalized-sort). \r\nDon't hesitate to ask any question on the data or the platform. Thanks for participating!\r\n\r\nAdam", "votes": null }, { "id": "115308", "postDate": "04/17/2016 22:22:20", "content": "Can I please get a clarification on the is_booking, time_stamp and cnt variables?
\n\n'Cnt' is defined as "Numer (sic) of similar events in the context of the same user session"
\n\n1) What is a similar event in the context refer to - can you give an example?
\n\n2) What does is_booking ==0 mean? I understand it is a click, but does that mean that a user is interacting with expedia.com (or related), sees a hotel in the search results they may be interested in and clicks to see more?
\n\n3) Are we only seeing a fraction of each user session log activity?
\n\n4) Take the first three records from the train set (user id 12). How do we make sense of this? This user clicked on 2 hotels of cluster 1 and then also booked 1 of cluster 1? Did they see three distinct hotels and finally chose one? It looks like they booked a hotel and then later in the session they clicked on another for the same check-in check-out but did not book?
", "rawMarkdown": "Can I please get a clarification on the is_booking, time_stamp and cnt variables?\r\n\r\n\r\n'Cnt' is defined as \"Numer (sic) of similar events in the context of the same user session\"\r\n\r\n1) What is a similar event in the context refer to - can you give an example? \r\n\r\n2) What does is_booking ==0 mean? I understand it is a click, but does that mean that a user is interacting with expedia.com (or related), sees a hotel in the search results they may be interested in and clicks to see more?\r\n\r\n3) Are we only seeing a fraction of each user session log activity?\r\n\r\n4) Take the first three records from the train set (user id 12). How do we make sense of this? This user clicked on 2 hotels of cluster 1 and then also booked 1 of cluster 1? Did they see three distinct hotels and finally chose one? It looks like they booked a hotel and then later in the session they clicked on another for the same check-in check-out but did not book?", "votes": null }, { "id": "115363", "postDate": "04/18/2016 11:24:42", "content": "Hello Adam\nWe have a problem about understanding how our submission entry should be. We assume that if we predict 5 hotel_cluster for every record in result set, only one of them is correct and the rest is wrong.is this assumption true or not? and if it's true how Mean Average Precision will be calculated for our result?(because in the examples it assumes that more than one prediction could be correct). if It's otherwise our assumption please explain the way of evaluation\nThanks
", "rawMarkdown": "Hello Adam\r\nWe have a problem about understanding how our submission entry should be. We assume that if we predict 5 hotel_cluster for every record in result set, only one of them is correct and the rest is wrong.is this assumption true or not? and if it's true how Mean Average Precision will be calculated for our result?(because in the examples it assumes that more than one prediction could be correct). if It's otherwise our assumption please explain the way of evaluation\r\nThanks", "votes": null }, { "id": "115369", "postDate": "04/18/2016 12:03:35", "content": "Hi,
\n\nmy understanding of MAP@5 is that if you predict the right cluster in the first spot you get a higher score than if you do so at position 5. But you still get a score there.
\n\nGerhard
", "rawMarkdown": "Hi,\r\n\r\nmy understanding of MAP@5 is that if you predict the right cluster in the first spot you get a higher score than if you do so at position 5. But you still get a score there.\r\n\r\nGerhard", "votes": null }, { "id": "115471", "postDate": "04/18/2016 19:46:29", "content": "@Inspector,
\n\n1) Events are only clicks and bookings, e.g. cnt=10 if is_booking=0 means that a user clicked 10 times on a link to a hotel infosite page within a user session. A user session is defined by user actions with inactivity less than 30 min.
\n\n2) is_booking = 0 means it's a click, e.g. a user sees a hotel in a search result page and clicks on it to see room rates, pictures, reviews, etc.
\n\n3) Correct. The data only contains bookings and clicks, other activities like hotel/flight/car searches, homepage visits, etc. are not considered here. On the other hand, the training data should contain all clicks and bookings for a given user.
\n\n4) "This user clicked on 2 hotels of cluster 1 and then also booked 1 of cluster 1?"\nCorrect.\n"Did they see three distinct hotels and finally chose one? "\nNo, the user clicked on one hotel, then booked one hotel, then he clicked on one hotel. Most likely these 3 rows correspond to the same hotel. There are 3 rows here because for the first 2 rows is_package = 1.\n"It looks like they booked a hotel and then later in the session they clicked on another for the same check-in check-out but did not book?"\nCorrect.
", "rawMarkdown": "Inspector,\r\n\r\n1) Events are only clicks and bookings, e.g. cnt=10 if is_booking=0 means that a user clicked 10 times on a link to a hotel infosite page within a user session. A user session is defined by user actions with inactivity less than 30 min.\r\n\r\n2) is_booking = 0 means it's a click, e.g. a user sees a hotel in a search result page and clicks on it to see room rates, pictures, reviews, etc.\r\n\r\n3) Correct. The data only contains bookings and clicks, other activities like hotel/flight/car searches, homepage visits, etc. are not considered here. On the other hand, the training data should contain all clicks and bookings for a given user.\r\n\r\n4) \"This user clicked on 2 hotels of cluster 1 and then also booked 1 of cluster 1?\"\r\nCorrect.\r\n\"Did they see three distinct hotels and finally chose one? \"\r\nNo, the user clicked on one hotel, then booked one hotel, then he clicked on one hotel. Most likely these 3 rows correspond to the same hotel. There are 3 rows here because for the first 2 rows is_package = 1.\r\n\"It looks like they booked a hotel and then later in the session they clicked on another for the same check-in check-out but did not book?\"\r\nCorrect.", "votes": null }, { "id": "115802", "postDate": "04/20/2016 08:22:09", "content": "Hi Adam, quick one for you..
\n\nMain Description: "...Expedia is challenging Kagglers to contextualize customer data and predict the likelihood a user will stay at 100 different hotel groups."
\n\nAre the clusters different hotel groups (As one particular chain of hotels) or are they clusters of similar hotels ?
\n\nie. clustered by distance from city center etc.
\n\nI think this would have a big influence on how to approach the problem.
\n\nThanks, \nDylan
", "rawMarkdown": "Hi Adam, quick one for you..\r\n\r\nMain Description: \"...Expedia is challenging Kagglers to contextualize customer data and predict the likelihood a user will stay at 100 different **hotel groups**.\"\r\n\r\nAre the clusters different hotel groups (As one particular chain of hotels) or are they clusters of similar hotels ? \r\n\r\nie. clustered by distance from city center etc.\r\n\r\nI think this would have a big influence on how to approach the problem.\r\n\r\nThanks, \r\nDylan", "votes": null }, { "id": "115805", "postDate": "04/20/2016 08:30:34", "content": "They are clusters of similar hotels. These clusters are based on hotels popularity, star rating, user review rating, price, distance from city center, amenities, etc. Hotels can belong to different clusters depending on seasonality (e.g. a hotel can be popular and expensive during summer, and cheap and not very popular during winter).
\n\nAdam
", "rawMarkdown": "They are clusters of similar hotels. These clusters are based on hotels popularity, star rating, user review rating, price, distance from city center, amenities, etc. Hotels can belong to different clusters depending on seasonality (e.g. a hotel can be popular and expensive during summer, and cheap and not very popular during winter).\r\n\r\nAdam", "votes": null }, { "id": "117185", "postDate": "04/27/2016 19:34:26", "content": "Hi Adam, thanks for the explanations.
\n\nCould you provide some info about the marketing channel ID ("channel" variable), thanks in advance.
\n\nRicardo
", "rawMarkdown": "Hi Adam, thanks for the explanations.\r\n\r\nCould you provide some info about the marketing channel ID (\"channel\" variable), thanks in advance.\r\n\r\nRicardo", "votes": null }, { "id": "117189", "postDate": "04/27/2016 19:44:22", "content": "Basically, it's an info on how a user got to the expedia web page. It can be something like: "Direct", "SEM" (e.g. via Google paid links), "Meta channel" (e.g. via Tripadvisor), etc.
", "rawMarkdown": "Basically, it's an info on how a user got to the expedia web page. It can be something like: \"Direct\", \"SEM\" (e.g. via Google paid links), \"Meta channel\" (e.g. via Tripadvisor), etc.", "votes": null }, { "id": "117307", "postDate": "04/28/2016 13:11:29", "content": "Hi, Adam.
\n\nI was exploring the dataset and noticed a strange thing about user_location_city and its relationship to user_location_country. I thought that any particular city should belong to one country only, but it appears that this is not the case. In fact, in the first million rows of the train.csv there is a city (26750) that appears in 17 different countries.
Could you please explain this? Could it be that every country has its own city ids that could overlap?
\n\nThanks,
\n\nDaria
", "rawMarkdown": "Hi, Adam.\r\n\r\nI was exploring the dataset and noticed a strange thing about `user_location_city` and its relationship to `user_location_country`. I thought that any particular city should belong to one country only, but it appears that this is not the case. In fact, in the first million rows of the train.csv there is a city (`26750`) that appears in 17 different countries.\r\n\r\nCould you please explain this? Could it be that every country has its own city ids that could overlap?\r\n\r\nThanks,\r\n\r\nDaria", "votes": null }, { "id": "117309", "postDate": "04/28/2016 13:23:46", "content": "If that is the only case then I would assume it is encoded label 'unknown' or something similar.
", "rawMarkdown": "If that is the only case then I would assume it is encoded label 'unknown' or something similar.", "votes": null }, { "id": "117313", "postDate": "04/28/2016 13:41:24", "content": "That was the most extreme example but there are 507 cities that have more than one country. About 187 thousand rows of the first million have one of these cities.
", "rawMarkdown": "That was the most extreme example but there are 507 cities that have more than one country. About 187 thousand rows of the first million have one of these cities.", "votes": null }, { "id": "117315", "postDate": "04/28/2016 13:43:58", "content": "Another reason might be that the encoding was done by name only, and there are plenty of examples where cities in two countries have the same name, but maybe better lets wait for Adam to write something about it.
", "rawMarkdown": "Another reason might be that the encoding was done by name only, and there are plenty of examples where cities in two countries have the same name, but maybe better lets wait for Adam to write something about it.", "votes": null }, { "id": "117317", "postDate": "04/28/2016 13:49:26", "content": "Correct, hashing was done by name only and there are cities with the same name in different countries.\nWhat it means is that user_location_city should be used only in the context of corresponding user_location_country/region, and user_location_region should be only used in the context of corresponding user_location_country.
", "rawMarkdown": "Correct, hashing was done by name only and there are cities with the same name in different countries.\r\nWhat it means is that user_location_city should be used only in the context of corresponding user_location_country/region, and user_location_region should be only used in the context of corresponding user_location_country.", "votes": null }, { "id": "117519", "postDate": "04/29/2016 14:03:30", "content": "@Adam, the same if valid for hotel_country / srch_destination_id and hotel_country / hotel_market? I notice that there are some srch_destination_id and hotel_market in more than one country.
Yes, it's possible for a market / srch_desitnation to span multiple countries. Also, it can happen that the underlying data is not correct (e.g. a hotel is assigned in our db to a wrong market).
", "rawMarkdown": "Yes, it's possible for a market / srch_desitnation to span multiple countries. Also, it can happen that the underlying data is not correct (e.g. a hotel is assigned in our db to a wrong market).", "votes": null }, { "id": "117748", "postDate": "04/30/2016 17:19:53", "content": "hey do we have to use destination.csv in our model as well?\nI'm new to all the competitions.. so please if you could elaborate in what needs to be done exactly?
", "rawMarkdown": "hey do we have to use destination.csv in our model as well?\r\nI'm new to all the competitions.. so please if you could elaborate in what needs to be done exactly?", "votes": null }, { "id": "117770", "postDate": "04/30/2016 20:32:51", "content": "You don't have to use it. If you get better scores without it. Why not?
\n\nGerhard
", "rawMarkdown": "You don't have to use it. If you get better scores without it. Why not?\r\n\r\nGerhard", "votes": null }, { "id": "117777", "postDate": "04/30/2016 21:19:25", "content": "Can anyone explain what is 'hotel_market' in the train file?
", "rawMarkdown": "Can anyone explain what is 'hotel_market' in the train file?", "votes": null }, { "id": "117832", "postDate": "05/01/2016 05:26:55", "content": "@Adam:
\n\nHi Adam
\n\nI work for Expedia India. Can i be a part of this competition as i didn't find any restrictions for Expedia employee.
", "rawMarkdown": "Adam:\r\n\r\nHi Adam\r\n\r\nI work for Expedia India. Can i be a part of this competition as i didn't find any restrictions for Expedia employee.", "votes": null }, { "id": "117834", "postDate": "05/01/2016 05:39:58", "content": "@Ashish
\n\nFrom the rules "Eligibility" section:
\n\n\n\n\nEmployees, officers and directors of Sponsor, Administrator, their\n respective successors, participating award suppliers and advertising\n and promotion agencies, judges, agents, and each of their respective\n subsidiaries, affiliates, agents, representatives, officers,\n directors, shareholders and employees (collectively, “Sponsor\n Parties”), and each of their immediate family members (spouse and\n parents, children and siblings, and their respective spouse,\n regardless of where they live) and individuals living in the same\n households of each of these individuals, whether related or not, of\n the above are not eligible to participate in the Competition.
\n
You're an employee of the Sponsor, so I think you're ineligible.
", "rawMarkdown": "Ashish\r\n\r\nFrom the rules \"Eligibility\" section:\r\n\r\n> Employees, officers and directors of Sponsor, Administrator, their\r\n> respective successors, participating award suppliers and advertising\r\n> and promotion agencies, judges, agents, and each of their respective\r\n> subsidiaries, affiliates, agents, representatives, officers,\r\n> directors, shareholders and employees (collectively, “Sponsor\r\n> Parties”), and each of their immediate family members (spouse and\r\n> parents, children and siblings, and their respective spouse,\r\n> regardless of where they live) and individuals living in the same\r\n> households of each of these individuals, whether related or not, of\r\n> the above are not eligible to participate in the Competition.\r\n\r\nYou're an employee of the Sponsor, so I think you're ineligible.", "votes": null }, { "id": "118314", "postDate": "05/03/2016 08:59:49", "content": "@Adam - one question regarding the orig_destination_distance.\nIn some cases it is NA. The data section explains, that it means for this particular case it was impossible to calculate the physical distance between user and the hotel. What could be the reasons for that?
", "rawMarkdown": "Adam - one question regarding the orig_destination_distance.\r\nIn some cases it is NA. The data section explains, that it means for this particular case it was impossible to calculate the physical distance between user and the hotel. What could be the reasons for that?", "votes": null }, { "id": "118316", "postDate": "05/03/2016 09:12:37", "content": "One reason is missing user location. Even if we know user location, we sometimes don't know lat-long of that location.
", "rawMarkdown": "One reason is missing user location. Even if we know user location, we sometimes don't know lat-long of that location.", "votes": null }, { "id": "118322", "postDate": "05/03/2016 09:40:01", "content": "Hi Adam,\nJust a confirmation : is the time contained in the variable 'date_time' a local time depending on the user location ?\nThank you
", "rawMarkdown": "Hi Adam,\r\nJust a confirmation : is the time contained in the variable 'date_time' a local time depending on the user location ?\r\nThank you", "votes": null }, { "id": "118323", "postDate": "05/03/2016 09:41:40", "content": "It's time relative to site_name.
", "rawMarkdown": "It's time relative to site_name.", "votes": null }, { "id": "118324", "postDate": "05/03/2016 09:47:43", "content": "Hi Adam. Could you please give us the distance unit?
", "rawMarkdown": "Hi Adam. Could you please give us the distance unit?", "votes": null }, { "id": "118325", "postDate": "05/03/2016 09:48:48", "content": "miles
", "rawMarkdown": "miles", "votes": null }, { "id": "118327", "postDate": "05/03/2016 09:57:03", "content": "Thank you
", "rawMarkdown": "Thank you", "votes": null }, { "id": "118465", "postDate": "05/03/2016 19:57:05", "content": "\n\n\nmiles
\n
What kind of distance is this? Flying distance (arc length accross the globe), driving distance or something else? Thanks.
", "rawMarkdown": "> miles\r\n\r\nWhat kind of distance is this? Flying distance (arc length accross the globe), driving distance or something else? Thanks.", "votes": null }, { "id": "118467", "postDate": "05/03/2016 19:58:35", "content": "Geodesic distance.
", "rawMarkdown": "Geodesic distance.", "votes": null }, { "id": "120107", "postDate": "05/15/2016 14:09:53", "content": "I am a little new to these contests. I submit 5 hotel clusters with an id.
\n\nIf one of those 5 hotel clusters actually got booked I then its counted as correct? All the user ids also booked something? I don't have to worry about figuring out if the user left the site before booking?
", "rawMarkdown": "I am a little new to these contests. I submit 5 hotel clusters with an id. \r\n\r\nIf one of those 5 hotel clusters actually got booked I then its counted as correct? All the user ids also booked something? I don't have to worry about figuring out if the user left the site before booking?", "votes": null }, { "id": "120387", "postDate": "05/17/2016 22:34:20", "content": "Hi Adam, does this means only when 'is_booking' == 0, the column 'cnt' can be a non-zero value, which means the user clicked several times during his session. And when 'is_booking' == 1, the 'cnt' must be 0?
", "rawMarkdown": "Hi Adam, does this means only when 'is_booking' == 0, the column 'cnt' can be a non-zero value, which means the user clicked several times during his session. And when 'is_booking' == 1, the 'cnt' must be 0?", "votes": null }, { "id": "120431", "postDate": "05/18/2016 06:47:41", "content": "Correct.
", "rawMarkdown": "Correct.", "votes": null }, { "id": "120437", "postDate": "05/18/2016 07:36:34", "content": "Hi Adam,
\n\nI'm sorry but I don't understand your answer.
\n\n'cnt' is never equal to 0 (booking event or not), but it may also be larger than 1 for a booking event. Are you saying that we should ignore 'cnt' for booking event ?
\n\nMay it be the number of clicks during the session where the booking event occured ?
", "rawMarkdown": "Hi Adam,\r\n\r\nI'm sorry but I don't understand your answer.\r\n\r\n'cnt' is never equal to 0 (booking event or not), but it may also be larger than 1 for a booking event. Are you saying that we should ignore 'cnt' for booking event ?\r\n\r\nMay it be the number of clicks during the session where the booking event occured ?", "votes": null }, { "id": "120441", "postDate": "05/18/2016 07:53:37", "content": "cnt is the # of clicks or booking for a particular hotel in a given user session. In the vast majority of cases there is only one booking for a given hotel per session, hence cnt == 1. In other words you can assume that if is_booking == 1 then cnt == 1.
", "rawMarkdown": "cnt is the # of clicks or booking for a particular hotel in a given user session. In the vast majority of cases there is only one booking for a given hotel per session, hence cnt == 1. In other words you can assume that if is_booking == 1 then cnt == 1.", "votes": null }, { "id": "120708", "postDate": "05/20/2016 02:11:49", "content": "@Adam\nHi, Adam.
\n\nI didn't understand exactly what "user event" field of output csv file.
\n\nWhat field is mapped within train set or test set? user id or destination id.. or etc.
\n\nThanks.
", "rawMarkdown": "Adam\r\nHi, Adam.\r\n\r\nI didn't understand exactly what \"user event\" field of output csv file.\r\n\r\nWhat field is mapped within train set or test set? user id or destination id.. or etc.\r\n\r\nThanks.", "votes": null }, { "id": "120740", "postDate": "05/20/2016 08:39:11", "content": "What is the "output csv file"?
", "rawMarkdown": "What is the \"output csv file\"?", "votes": null }, { "id": "120744", "postDate": "05/20/2016 09:20:33", "content": "Sorry, I confused you. I mean submission csv file.
", "rawMarkdown": "Sorry, I confused you. I mean submission csv file.", "votes": null }, { "id": "120758", "postDate": "05/20/2016 11:10:06", "content": "If you're referring to the 1st column in the submission file, it's just the record identifier.
", "rawMarkdown": "If you're referring to the 1st column in the submission file, it's just the record identifier.", "votes": null }, { "id": "564306", "postDate": "06/29/2019 09:05:03", "content": "Can you please provide the labels for test set instances?
", "rawMarkdown": "Can you please provide the labels for test set instances?", "votes": null } ], "comments": [ { "id": 115308, "author_name": "inspector", "author_url": "", "post_date": "04/17/2016 22:22:20", "content": "Can I please get a clarification on the is_booking, time_stamp and cnt variables?
\n\n'Cnt' is defined as "Numer (sic) of similar events in the context of the same user session"
\n\n1) What is a similar event in the context refer to - can you give an example?
\n\n2) What does is_booking ==0 mean? I understand it is a click, but does that mean that a user is interacting with expedia.com (or related), sees a hotel in the search results they may be interested in and clicks to see more?
\n\n3) Are we only seeing a fraction of each user session log activity?
\n\n4) Take the first three records from the train set (user id 12). How do we make sense of this? This user clicked on 2 hotels of cluster 1 and then also booked 1 of cluster 1? Did they see three distinct hotels and finally chose one? It looks like they booked a hotel and then later in the session they clicked on another for the same check-in check-out but did not book?
", "votes": null, "replies": [] }, { "id": 115363, "author_name": "amin0ooo0", "author_url": "", "post_date": "04/18/2016 11:24:42", "content": "Hello Adam\nWe have a problem about understanding how our submission entry should be. We assume that if we predict 5 hotel_cluster for every record in result set, only one of them is correct and the rest is wrong.is this assumption true or not? and if it's true how Mean Average Precision will be calculated for our result?(because in the examples it assumes that more than one prediction could be correct). if It's otherwise our assumption please explain the way of evaluation\nThanks
", "votes": null, "replies": [] }, { "id": 115369, "author_name": "mightybird", "author_url": "", "post_date": "04/18/2016 12:03:35", "content": "Hi,
\n\nmy understanding of MAP@5 is that if you predict the right cluster in the first spot you get a higher score than if you do so at position 5. But you still get a score there.
\n\nGerhard
", "votes": null, "replies": [] }, { "id": 115471, "author_name": "adamwoz", "author_url": "", "post_date": "04/18/2016 19:46:29", "content": "@Inspector,
\n\n1) Events are only clicks and bookings, e.g. cnt=10 if is_booking=0 means that a user clicked 10 times on a link to a hotel infosite page within a user session. A user session is defined by user actions with inactivity less than 30 min.
\n\n2) is_booking = 0 means it's a click, e.g. a user sees a hotel in a search result page and clicks on it to see room rates, pictures, reviews, etc.
\n\n3) Correct. The data only contains bookings and clicks, other activities like hotel/flight/car searches, homepage visits, etc. are not considered here. On the other hand, the training data should contain all clicks and bookings for a given user.
\n\n4) "This user clicked on 2 hotels of cluster 1 and then also booked 1 of cluster 1?"\nCorrect.\n"Did they see three distinct hotels and finally chose one? "\nNo, the user clicked on one hotel, then booked one hotel, then he clicked on one hotel. Most likely these 3 rows correspond to the same hotel. There are 3 rows here because for the first 2 rows is_package = 1.\n"It looks like they booked a hotel and then later in the session they clicked on another for the same check-in check-out but did not book?"\nCorrect.
", "votes": null, "replies": [] }, { "id": 115802, "author_name": "kellybus", "author_url": "", "post_date": "04/20/2016 08:22:09", "content": "Hi Adam, quick one for you..
\n\nMain Description: "...Expedia is challenging Kagglers to contextualize customer data and predict the likelihood a user will stay at 100 different hotel groups."
\n\nAre the clusters different hotel groups (As one particular chain of hotels) or are they clusters of similar hotels ?
\n\nie. clustered by distance from city center etc.
\n\nI think this would have a big influence on how to approach the problem.
\n\nThanks, \nDylan
", "votes": null, "replies": [] }, { "id": 115805, "author_name": "adamwoz", "author_url": "", "post_date": "04/20/2016 08:30:34", "content": "They are clusters of similar hotels. These clusters are based on hotels popularity, star rating, user review rating, price, distance from city center, amenities, etc. Hotels can belong to different clusters depending on seasonality (e.g. a hotel can be popular and expensive during summer, and cheap and not very popular during winter).
\n\nAdam
", "votes": null, "replies": [] }, { "id": 117185, "author_name": "ricardom", "author_url": "", "post_date": "04/27/2016 19:34:26", "content": "Hi Adam, thanks for the explanations.
\n\nCould you provide some info about the marketing channel ID ("channel" variable), thanks in advance.
\n\nRicardo
", "votes": null, "replies": [] }, { "id": 117189, "author_name": "adamwoz", "author_url": "", "post_date": "04/27/2016 19:44:22", "content": "Basically, it's an info on how a user got to the expedia web page. It can be something like: "Direct", "SEM" (e.g. via Google paid links), "Meta channel" (e.g. via Tripadvisor), etc.
", "votes": null, "replies": [] }, { "id": 117307, "author_name": "dvasyukova", "author_url": "", "post_date": "04/28/2016 13:11:29", "content": "Hi, Adam.
\n\nI was exploring the dataset and noticed a strange thing about user_location_city and its relationship to user_location_country. I thought that any particular city should belong to one country only, but it appears that this is not the case. In fact, in the first million rows of the train.csv there is a city (26750) that appears in 17 different countries.
Could you please explain this? Could it be that every country has its own city ids that could overlap?
\n\nThanks,
\n\nDaria
", "votes": null, "replies": [] }, { "id": 117309, "author_name": "mpekalski", "author_url": "", "post_date": "04/28/2016 13:23:46", "content": "If that is the only case then I would assume it is encoded label 'unknown' or something similar.
", "votes": null, "replies": [] }, { "id": 117313, "author_name": "dvasyukova", "author_url": "", "post_date": "04/28/2016 13:41:24", "content": "That was the most extreme example but there are 507 cities that have more than one country. About 187 thousand rows of the first million have one of these cities.
", "votes": null, "replies": [] }, { "id": 117315, "author_name": "mpekalski", "author_url": "", "post_date": "04/28/2016 13:43:58", "content": "Another reason might be that the encoding was done by name only, and there are plenty of examples where cities in two countries have the same name, but maybe better lets wait for Adam to write something about it.
", "votes": null, "replies": [] }, { "id": 117317, "author_name": "adamwoz", "author_url": "", "post_date": "04/28/2016 13:49:26", "content": "Correct, hashing was done by name only and there are cities with the same name in different countries.\nWhat it means is that user_location_city should be used only in the context of corresponding user_location_country/region, and user_location_region should be only used in the context of corresponding user_location_country.
", "votes": null, "replies": [] }, { "id": 117519, "author_name": "bguberfain", "author_url": "", "post_date": "04/29/2016 14:03:30", "content": "@Adam, the same if valid for hotel_country / srch_destination_id and hotel_country / hotel_market? I notice that there are some srch_destination_id and hotel_market in more than one country.
Yes, it's possible for a market / srch_desitnation to span multiple countries. Also, it can happen that the underlying data is not correct (e.g. a hotel is assigned in our db to a wrong market).
", "votes": null, "replies": [] }, { "id": 117748, "author_name": "purvajapatil", "author_url": "", "post_date": "04/30/2016 17:19:53", "content": "hey do we have to use destination.csv in our model as well?\nI'm new to all the competitions.. so please if you could elaborate in what needs to be done exactly?
", "votes": null, "replies": [] }, { "id": 117770, "author_name": "mightybird", "author_url": "", "post_date": "04/30/2016 20:32:51", "content": "You don't have to use it. If you get better scores without it. Why not?
\n\nGerhard
", "votes": null, "replies": [] }, { "id": 117777, "author_name": "zhenjuan", "author_url": "", "post_date": "04/30/2016 21:19:25", "content": "Can anyone explain what is 'hotel_market' in the train file?
", "votes": null, "replies": [] }, { "id": 117832, "author_name": "ashishagg", "author_url": "", "post_date": "05/01/2016 05:26:55", "content": "@Adam:
\n\nHi Adam
\n\nI work for Expedia India. Can i be a part of this competition as i didn't find any restrictions for Expedia employee.
", "votes": null, "replies": [] }, { "id": 117834, "author_name": "brandenkmurray", "author_url": "", "post_date": "05/01/2016 05:39:58", "content": "@Ashish
\n\nFrom the rules "Eligibility" section:
\n\n\n\n\nEmployees, officers and directors of Sponsor, Administrator, their\n respective successors, participating award suppliers and advertising\n and promotion agencies, judges, agents, and each of their respective\n subsidiaries, affiliates, agents, representatives, officers,\n directors, shareholders and employees (collectively, “Sponsor\n Parties”), and each of their immediate family members (spouse and\n parents, children and siblings, and their respective spouse,\n regardless of where they live) and individuals living in the same\n households of each of these individuals, whether related or not, of\n the above are not eligible to participate in the Competition.
\n
You're an employee of the Sponsor, so I think you're ineligible.
", "votes": null, "replies": [] }, { "id": 118314, "author_name": "narsil", "author_url": "", "post_date": "05/03/2016 08:59:49", "content": "@Adam - one question regarding the orig_destination_distance.\nIn some cases it is NA. The data section explains, that it means for this particular case it was impossible to calculate the physical distance between user and the hotel. What could be the reasons for that?
", "votes": null, "replies": [] }, { "id": 118316, "author_name": "adamwoz", "author_url": "", "post_date": "05/03/2016 09:12:37", "content": "One reason is missing user location. Even if we know user location, we sometimes don't know lat-long of that location.
", "votes": null, "replies": [] }, { "id": 118322, "author_name": "amine2684", "author_url": "", "post_date": "05/03/2016 09:40:01", "content": "Hi Adam,\nJust a confirmation : is the time contained in the variable 'date_time' a local time depending on the user location ?\nThank you
", "votes": null, "replies": [] }, { "id": 118323, "author_name": "adamwoz", "author_url": "", "post_date": "05/03/2016 09:41:40", "content": "It's time relative to site_name.
", "votes": null, "replies": [] }, { "id": 118324, "author_name": "santiagomota", "author_url": "", "post_date": "05/03/2016 09:47:43", "content": "Hi Adam. Could you please give us the distance unit?
", "votes": null, "replies": [] }, { "id": 118325, "author_name": "adamwoz", "author_url": "", "post_date": "05/03/2016 09:48:48", "content": "miles
", "votes": null, "replies": [] }, { "id": 118327, "author_name": "santiagomota", "author_url": "", "post_date": "05/03/2016 09:57:03", "content": "Thank you
", "votes": null, "replies": [] }, { "id": 118465, "author_name": "vspinu", "author_url": "", "post_date": "05/03/2016 19:57:05", "content": "\n\n\nmiles
\n
What kind of distance is this? Flying distance (arc length accross the globe), driving distance or something else? Thanks.
", "votes": null, "replies": [] }, { "id": 118467, "author_name": "adamwoz", "author_url": "", "post_date": "05/03/2016 19:58:35", "content": "Geodesic distance.
", "votes": null, "replies": [] }, { "id": 120107, "author_name": "illidari", "author_url": "", "post_date": "05/15/2016 14:09:53", "content": "I am a little new to these contests. I submit 5 hotel clusters with an id.
\n\nIf one of those 5 hotel clusters actually got booked I then its counted as correct? All the user ids also booked something? I don't have to worry about figuring out if the user left the site before booking?
", "votes": null, "replies": [] }, { "id": 120387, "author_name": "jiaxit", "author_url": "", "post_date": "05/17/2016 22:34:20", "content": "Hi Adam, does this means only when 'is_booking' == 0, the column 'cnt' can be a non-zero value, which means the user clicked several times during his session. And when 'is_booking' == 1, the 'cnt' must be 0?
", "votes": null, "replies": [] }, { "id": 120431, "author_name": "adamwoz", "author_url": "", "post_date": "05/18/2016 06:47:41", "content": "Correct.
", "votes": null, "replies": [] }, { "id": 120437, "author_name": "fsimond", "author_url": "", "post_date": "05/18/2016 07:36:34", "content": "Hi Adam,
\n\nI'm sorry but I don't understand your answer.
\n\n'cnt' is never equal to 0 (booking event or not), but it may also be larger than 1 for a booking event. Are you saying that we should ignore 'cnt' for booking event ?
\n\nMay it be the number of clicks during the session where the booking event occured ?
", "votes": null, "replies": [] }, { "id": 120441, "author_name": "adamwoz", "author_url": "", "post_date": "05/18/2016 07:53:37", "content": "cnt is the # of clicks or booking for a particular hotel in a given user session. In the vast majority of cases there is only one booking for a given hotel per session, hence cnt == 1. In other words you can assume that if is_booking == 1 then cnt == 1.
", "votes": null, "replies": [] }, { "id": 120708, "author_name": "luvsword", "author_url": "", "post_date": "05/20/2016 02:11:49", "content": "@Adam\nHi, Adam.
\n\nI didn't understand exactly what "user event" field of output csv file.
\n\nWhat field is mapped within train set or test set? user id or destination id.. or etc.
\n\nThanks.
", "votes": null, "replies": [] }, { "id": 120740, "author_name": "adamwoz", "author_url": "", "post_date": "05/20/2016 08:39:11", "content": "What is the "output csv file"?
", "votes": null, "replies": [] }, { "id": 120744, "author_name": "luvsword", "author_url": "", "post_date": "05/20/2016 09:20:33", "content": "Sorry, I confused you. I mean submission csv file.
", "votes": null, "replies": [] }, { "id": 120758, "author_name": "adamwoz", "author_url": "", "post_date": "05/20/2016 11:10:06", "content": "If you're referring to the 1st column in the submission file, it's just the record identifier.
", "votes": null, "replies": [] }, { "id": 564306, "author_name": "jwdsoft", "author_url": "", "post_date": "06/29/2019 09:05:03", "content": "Can you please provide the labels for test set instances?
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "115037": "Welcome to the Expedia Hotel Recommendations competition. The goal is to predict which class of hotel was purchased given past purchases and clicks. This \r\nis the second time we partner with Kaggle to solve interesting problems (the first competition was https://www.kaggle.com/c/expedia-personalized-sort). \r\nDon't hesitate to ask any question on the data or the platform. Thanks for participating!\r\n\r\nAdam", "115308": "Can I please get a clarification on the is_booking, time_stamp and cnt variables?\r\n\r\n\r\n'Cnt' is defined as \"Numer (sic) of similar events in the context of the same user session\"\r\n\r\n1) What is a similar event in the context refer to - can you give an example? \r\n\r\n2) What does is_booking ==0 mean? I understand it is a click, but does that mean that a user is interacting with expedia.com (or related), sees a hotel in the search results they may be interested in and clicks to see more?\r\n\r\n3) Are we only seeing a fraction of each user session log activity?\r\n\r\n4) Take the first three records from the train set (user id 12). How do we make sense of this? This user clicked on 2 hotels of cluster 1 and then also booked 1 of cluster 1? Did they see three distinct hotels and finally chose one? It looks like they booked a hotel and then later in the session they clicked on another for the same check-in check-out but did not book?", "115363": "Hello Adam\r\nWe have a problem about understanding how our submission entry should be. We assume that if we predict 5 hotel_cluster for every record in result set, only one of them is correct and the rest is wrong.is this assumption true or not? and if it's true how Mean Average Precision will be calculated for our result?(because in the examples it assumes that more than one prediction could be correct). if It's otherwise our assumption please explain the way of evaluation\r\nThanks", "115369": "Hi,\r\n\r\nmy understanding of MAP@5 is that if you predict the right cluster in the first spot you get a higher score than if you do so at position 5. But you still get a score there.\r\n\r\nGerhard", "115471": "Inspector,\r\n\r\n1) Events are only clicks and bookings, e.g. cnt=10 if is_booking=0 means that a user clicked 10 times on a link to a hotel infosite page within a user session. A user session is defined by user actions with inactivity less than 30 min.\r\n\r\n2) is_booking = 0 means it's a click, e.g. a user sees a hotel in a search result page and clicks on it to see room rates, pictures, reviews, etc.\r\n\r\n3) Correct. The data only contains bookings and clicks, other activities like hotel/flight/car searches, homepage visits, etc. are not considered here. On the other hand, the training data should contain all clicks and bookings for a given user.\r\n\r\n4) \"This user clicked on 2 hotels of cluster 1 and then also booked 1 of cluster 1?\"\r\nCorrect.\r\n\"Did they see three distinct hotels and finally chose one? \"\r\nNo, the user clicked on one hotel, then booked one hotel, then he clicked on one hotel. Most likely these 3 rows correspond to the same hotel. There are 3 rows here because for the first 2 rows is_package = 1.\r\n\"It looks like they booked a hotel and then later in the session they clicked on another for the same check-in check-out but did not book?\"\r\nCorrect.", "115802": "Hi Adam, quick one for you..\r\n\r\nMain Description: \"...Expedia is challenging Kagglers to contextualize customer data and predict the likelihood a user will stay at 100 different **hotel groups**.\"\r\n\r\nAre the clusters different hotel groups (As one particular chain of hotels) or are they clusters of similar hotels ? \r\n\r\nie. clustered by distance from city center etc.\r\n\r\nI think this would have a big influence on how to approach the problem.\r\n\r\nThanks, \r\nDylan", "115805": "They are clusters of similar hotels. These clusters are based on hotels popularity, star rating, user review rating, price, distance from city center, amenities, etc. Hotels can belong to different clusters depending on seasonality (e.g. a hotel can be popular and expensive during summer, and cheap and not very popular during winter).\r\n\r\nAdam", "117185": "Hi Adam, thanks for the explanations.\r\n\r\nCould you provide some info about the marketing channel ID (\"channel\" variable), thanks in advance.\r\n\r\nRicardo", "117189": "Basically, it's an info on how a user got to the expedia web page. It can be something like: \"Direct\", \"SEM\" (e.g. via Google paid links), \"Meta channel\" (e.g. via Tripadvisor), etc.", "117307": "Hi, Adam.\r\n\r\nI was exploring the dataset and noticed a strange thing about `user_location_city` and its relationship to `user_location_country`. I thought that any particular city should belong to one country only, but it appears that this is not the case. In fact, in the first million rows of the train.csv there is a city (`26750`) that appears in 17 different countries.\r\n\r\nCould you please explain this? Could it be that every country has its own city ids that could overlap?\r\n\r\nThanks,\r\n\r\nDaria", "117309": "If that is the only case then I would assume it is encoded label 'unknown' or something similar.", "117313": "That was the most extreme example but there are 507 cities that have more than one country. About 187 thousand rows of the first million have one of these cities.", "117315": "Another reason might be that the encoding was done by name only, and there are plenty of examples where cities in two countries have the same name, but maybe better lets wait for Adam to write something about it.", "117317": "Correct, hashing was done by name only and there are cities with the same name in different countries.\r\nWhat it means is that user_location_city should be used only in the context of corresponding user_location_country/region, and user_location_region should be only used in the context of corresponding user_location_country.", "117519": "Adam, the same if valid for `hotel_country` / `srch_destination_id` and `hotel_country` / `hotel_market`? I notice that there are some `srch_destination_id` and `hotel_market` in more than one country.", "117530": "Yes, it's possible for a market / srch_desitnation to span multiple countries. Also, it can happen that the underlying data is not correct (e.g. a hotel is assigned in our db to a wrong market).", "117748": "hey do we have to use destination.csv in our model as well?\r\nI'm new to all the competitions.. so please if you could elaborate in what needs to be done exactly?", "117770": "You don't have to use it. If you get better scores without it. Why not?\r\n\r\nGerhard", "117777": "Can anyone explain what is 'hotel_market' in the train file?", "117832": "Adam:\r\n\r\nHi Adam\r\n\r\nI work for Expedia India. Can i be a part of this competition as i didn't find any restrictions for Expedia employee.", "117834": "Ashish\r\n\r\nFrom the rules \"Eligibility\" section:\r\n\r\n> Employees, officers and directors of Sponsor, Administrator, their\r\n> respective successors, participating award suppliers and advertising\r\n> and promotion agencies, judges, agents, and each of their respective\r\n> subsidiaries, affiliates, agents, representatives, officers,\r\n> directors, shareholders and employees (collectively, “Sponsor\r\n> Parties”), and each of their immediate family members (spouse and\r\n> parents, children and siblings, and their respective spouse,\r\n> regardless of where they live) and individuals living in the same\r\n> households of each of these individuals, whether related or not, of\r\n> the above are not eligible to participate in the Competition.\r\n\r\nYou're an employee of the Sponsor, so I think you're ineligible.", "118314": "Adam - one question regarding the orig_destination_distance.\r\nIn some cases it is NA. The data section explains, that it means for this particular case it was impossible to calculate the physical distance between user and the hotel. What could be the reasons for that?", "118316": "One reason is missing user location. Even if we know user location, we sometimes don't know lat-long of that location.", "118322": "Hi Adam,\r\nJust a confirmation : is the time contained in the variable 'date_time' a local time depending on the user location ?\r\nThank you", "118323": "It's time relative to site_name.", "118324": "Hi Adam. Could you please give us the distance unit?", "118325": "miles", "118327": "Thank you", "118465": "> miles\r\n\r\nWhat kind of distance is this? Flying distance (arc length accross the globe), driving distance or something else? Thanks.", "118467": "Geodesic distance.", "120107": "I am a little new to these contests. I submit 5 hotel clusters with an id. \r\n\r\nIf one of those 5 hotel clusters actually got booked I then its counted as correct? All the user ids also booked something? I don't have to worry about figuring out if the user left the site before booking?", "120387": "Hi Adam, does this means only when 'is_booking' == 0, the column 'cnt' can be a non-zero value, which means the user clicked several times during his session. And when 'is_booking' == 1, the 'cnt' must be 0?", "120431": "Correct.", "120437": "Hi Adam,\r\n\r\nI'm sorry but I don't understand your answer.\r\n\r\n'cnt' is never equal to 0 (booking event or not), but it may also be larger than 1 for a booking event. Are you saying that we should ignore 'cnt' for booking event ?\r\n\r\nMay it be the number of clicks during the session where the booking event occured ?", "120441": "cnt is the # of clicks or booking for a particular hotel in a given user session. In the vast majority of cases there is only one booking for a given hotel per session, hence cnt == 1. In other words you can assume that if is_booking == 1 then cnt == 1.", "120708": "Adam\r\nHi, Adam.\r\n\r\nI didn't understand exactly what \"user event\" field of output csv file.\r\n\r\nWhat field is mapped within train set or test set? user id or destination id.. or etc.\r\n\r\nThanks.", "120740": "What is the \"output csv file\"?", "120744": "Sorry, I confused you. I mean submission csv file.", "120758": "If you're referring to the 1st column in the submission file, it's just the record identifier.", "564306": "Can you please provide the labels for test set instances?" }, "source": "meta" }