I have to admit that my initial plan was to use a classification algorithm for this data, but then I saw that most of the competitors use regression, which seems to be much better. Is it always the case that regression is a better approach for an ordinal dependent variable? Or should we always try both regression and classification to see which one is more suitable in a specific scenario?
", "messages": [ { "id": "3005033", "postDate": "10/02/2024 13:42:55", "content": "I have to admit that my initial plan was to use a classification algorithm for this data, but then I saw that most of the competitors use regression, which seems to be much better. Is it always the case that regression is a better approach for an ordinal dependent variable? Or should we always try both regression and classification to see which one is more suitable in a specific scenario?
", "rawMarkdown": "I have to admit that my initial plan was to use a classification algorithm for this data, but then I saw that most of the competitors use regression, which seems to be much better. Is it always the case that regression is a better approach for an ordinal dependent variable? Or should we always try both regression and classification to see which one is more suitable in a specific scenario?", "votes": null }, { "id": "3006254", "postDate": "10/03/2024 23:47:30", "content": "You can find a good answer to you question here
\nhttps://stats.stackexchange.com/questions/493254/why-ordinal-target-in-classification-problems-need-special-attention
TLDR; you can think of the problem as an 'ordinal regression task' due to the inherent ordering in the 'sii' variable that we are trying to predict.
", "rawMarkdown": "You can find a good answer to you question here\nhttps://stats.stackexchange.com/questions/493254/why-ordinal-target-in-classification-problems-need-special-attention\n\nTLDR; you can think of the problem as an 'ordinal regression task' due to the inherent ordering in the 'sii' variable that we are trying to predict.", "votes": null }, { "id": "3025825", "postDate": "10/23/2024 07:03:05", "content": "\n", "rawMarkdown": ">In the case of classification, the order of classes is not considered as you can see the loss function.\n>However in this competition, there is an ordinal relationship between class labels.\n>Such a problem setting is generally referred to as ordinal regression.\nhttps://www.kaggle.com/competitions/child-mind-institute-problematic-internet-use/discussion/539949#3014837", "votes": null } ], "comments": [ { "id": 3006254, "author_name": "rafidmahbub", "author_url": "", "post_date": "10/03/2024 23:47:30", "content": "In the case of classification, the order of classes is not considered as you can see the loss function.
\n
\n However in this competition, there is an ordinal relationship between class labels.
\n Such a problem setting is generally referred to as ordinal regression.
\n https://www.kaggle.com/competitions/child-mind-institute-problematic-internet-use/discussion/539949#3014837
You can find a good answer to you question here
\nhttps://stats.stackexchange.com/questions/493254/why-ordinal-target-in-classification-problems-need-special-attention
TLDR; you can think of the problem as an 'ordinal regression task' due to the inherent ordering in the 'sii' variable that we are trying to predict.
", "votes": null, "replies": [] }, { "id": 3025825, "author_name": "clearwaterkzk", "author_url": "", "post_date": "10/23/2024 07:03:05", "content": "\n", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "3005033": "I have to admit that my initial plan was to use a classification algorithm for this data, but then I saw that most of the competitors use regression, which seems to be much better. Is it always the case that regression is a better approach for an ordinal dependent variable? Or should we always try both regression and classification to see which one is more suitable in a specific scenario?", "3006254": "You can find a good answer to you question here\nhttps://stats.stackexchange.com/questions/493254/why-ordinal-target-in-classification-problems-need-special-attention\n\nTLDR; you can think of the problem as an 'ordinal regression task' due to the inherent ordering in the 'sii' variable that we are trying to predict.", "3025825": ">In the case of classification, the order of classes is not considered as you can see the loss function.\n>However in this competition, there is an ordinal relationship between class labels.\n>Such a problem setting is generally referred to as ordinal regression.\nhttps://www.kaggle.com/competitions/child-mind-institute-problematic-internet-use/discussion/539949#3014837" }, "source": "meta" }In the case of classification, the order of classes is not considered as you can see the loss function.
\n
\n However in this competition, there is an ordinal relationship between class labels.
\n Such a problem setting is generally referred to as ordinal regression.
\n https://www.kaggle.com/competitions/child-mind-institute-problematic-internet-use/discussion/539949#3014837