{ "id": 20914, "title": "Number of predictions required for each row_id?", "url": "/competitions/facebook-v-predicting-check-ins/discussion/20914", "author_name": "Albert Lee", "post_date": "2016-05-12T20:07:32.227000", "votes": 0, "comment_count": 5, "views": 945, "content": "

I see that the sample submission looks like following:

\n\n
row_id,place_id\n0,3073560757 9004412889 5652080691\n1,1652178628 4379515211 6612350960\n2,4894407065 3920195083 7608574746\n3,7272466660 2004687925 1455486822\n4,4720452725 4967325204 4782917866\n..\n
\n\n

That is, each row has three place_id(s)

\n\n

Questions:

\n\n

1) Do we have to provide three predictions for each row_id? or can it have more?

\n\n

2) Does the order of place_ids matter?

", "messages": [ { "id": 119847, "postDate": "2016-05-13T04:02:35.203Z", "content": "

@Jared Turkewitz: every row_id in the test set corresponds to exactly one place_id in the train set

", "rawMarkdown": "@Jared Turkewitz: every row_id in the test set corresponds to exactly one place_id in the train set\r\n", "votes": 3 }, { "id": 119798, "postDate": "2016-05-12T20:23:44.240Z", "content": "

From: metric you can get the answers to your question. In short there is never any reason to not just submit exactly 3 different predictions, and you should order them by descending confidence in your prediction (most probable place_id should be 1st).

\n\n

I have a question about the test set - do all row_ids in test set correspond to a place_id in the train set? In other words if we made perfect predictions would we get a score of 1? Also is it true that each row_id in test corresponds to exactly one place_id (this seems to be the case for train)?

", "rawMarkdown": "From: [metric](https://www.kaggle.com/c/facebook-v-predicting-check-ins/details/evaluation) you can get the answers to your question. In short there is never any reason to not just submit exactly 3 different predictions, and you should order them by descending confidence in your prediction (most probable place_id should be 1st).\r\n\r\nI have a question about the test set - do all row_ids in test set correspond to a place_id in the train set? In other words if we made perfect predictions would we get a score of 1? Also is it true that each row_id in test corresponds to exactly one place_id (this seems to be the case for train)? \r\n", "votes": 3 }, { "id": 119799, "postDate": "2016-05-12T20:23:59.117Z", "content": "

Evaluation page says:\nYou may submit up to 3 predictions for each check in. (https://www.kaggle.com/c/facebook-v-predicting-check-ins/details/evaluation)

\n\n

Looks like it's an answer on first question.

", "rawMarkdown": "Evaluation page says:\r\nYou may submit **up to 3** predictions for each check in. (https://www.kaggle.com/c/facebook-v-predicting-check-ins/details/evaluation)\r\n\r\nLooks like it's an answer on first question.\r\n", "votes": 1 }, { "id": 121255, "postDate": "2016-05-25T07:09:43.077Z", "content": "

I am not sure, which math model predicts more than one recommendation!!! How do we generate 3 priority outputs then ?

", "rawMarkdown": "I am not sure, which math model predicts more than one recommendation!!! How do we generate 3 priority outputs then ? ", "votes": -2 }, { "id": 119793, "postDate": "2016-05-12T20:07:32.227Z", "content": "

I see that the sample submission looks like following:

\n\n
row_id,place_id\n0,3073560757 9004412889 5652080691\n1,1652178628 4379515211 6612350960\n2,4894407065 3920195083 7608574746\n3,7272466660 2004687925 1455486822\n4,4720452725 4967325204 4782917866\n..\n
\n\n

That is, each row has three place_id(s)

\n\n

Questions:

\n\n

1) Do we have to provide three predictions for each row_id? or can it have more?

\n\n

2) Does the order of place_ids matter?

", "rawMarkdown": "I see that the sample submission looks like following:\r\n\r\n row_id,place_id\r\n 0,3073560757 9004412889 5652080691\r\n 1,1652178628 4379515211 6612350960\r\n 2,4894407065 3920195083 7608574746\r\n 3,7272466660 2004687925 1455486822\r\n 4,4720452725 4967325204 4782917866\r\n ..\r\n\r\nThat is, each row has three place_id(s)\r\n\r\nQuestions:\r\n\r\n1) Do we have to provide three predictions for each row_id? or can it have more?\r\n\r\n2) Does the order of place_ids matter?" }, { "id": 119817, "postDate": "2016-05-12T21:20:35.927Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 119847, "author_name": "Anaconda", "author_url": "", "post_date": "2016-05-13T04:02:35.203000", "content": "

@Jared Turkewitz: every row_id in the test set corresponds to exactly one place_id in the train set

", "votes": 3, "replies": [] }, { "id": 119798, "author_name": "Jared Turkewitz", "author_url": "", "post_date": "2016-05-12T20:23:44.240000", "content": "

From: metric you can get the answers to your question. In short there is never any reason to not just submit exactly 3 different predictions, and you should order them by descending confidence in your prediction (most probable place_id should be 1st).

\n\n

I have a question about the test set - do all row_ids in test set correspond to a place_id in the train set? In other words if we made perfect predictions would we get a score of 1? Also is it true that each row_id in test corresponds to exactly one place_id (this seems to be the case for train)?

", "votes": 3, "replies": [] }, { "id": 119799, "author_name": "Vasily Zverkov", "author_url": "", "post_date": "2016-05-12T20:23:59.117000", "content": "

Evaluation page says:\nYou may submit up to 3 predictions for each check in. (https://www.kaggle.com/c/facebook-v-predicting-check-ins/details/evaluation)

\n\n

Looks like it's an answer on first question.

", "votes": 1, "replies": [] }, { "id": 121255, "author_name": "SandeepKarmarkar", "author_url": "", "post_date": "2016-05-25T07:09:43.077000", "content": "

I am not sure, which math model predicts more than one recommendation!!! How do we generate 3 priority outputs then ?

", "votes": -2, "replies": [] }, { "id": 119817, "author_name": "", "author_url": "", "post_date": "2016-05-12T21:20:35.927000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "119847": "@Jared Turkewitz: every row_id in the test set corresponds to exactly one place_id in the train set\r\n", "119798": "From: [metric](https://www.kaggle.com/c/facebook-v-predicting-check-ins/details/evaluation) you can get the answers to your question. In short there is never any reason to not just submit exactly 3 different predictions, and you should order them by descending confidence in your prediction (most probable place_id should be 1st).\r\n\r\nI have a question about the test set - do all row_ids in test set correspond to a place_id in the train set? In other words if we made perfect predictions would we get a score of 1? Also is it true that each row_id in test corresponds to exactly one place_id (this seems to be the case for train)? \r\n", "119799": "Evaluation page says:\r\nYou may submit **up to 3** predictions for each check in. (https://www.kaggle.com/c/facebook-v-predicting-check-ins/details/evaluation)\r\n\r\nLooks like it's an answer on first question.\r\n", "121255": "I am not sure, which math model predicts more than one recommendation!!! How do we generate 3 priority outputs then ? ", "119793": "I see that the sample submission looks like following:\r\n\r\n row_id,place_id\r\n 0,3073560757 9004412889 5652080691\r\n 1,1652178628 4379515211 6612350960\r\n 2,4894407065 3920195083 7608574746\r\n 3,7272466660 2004687925 1455486822\r\n 4,4720452725 4967325204 4782917866\r\n ..\r\n\r\nThat is, each row has three place_id(s)\r\n\r\nQuestions:\r\n\r\n1) Do we have to provide three predictions for each row_id? or can it have more?\r\n\r\n2) Does the order of place_ids matter?", "119817": "" } }