I looked at the documentation of sklearn.metrics.roc_auc_score , i saw that there is an argument multi_class which needs to be passed as either 'ovo' or 'ovr' based on what kind of roc_auc we want for the case of multi class classfication otherwise it raises an error , as i checked in the metric given by kaggle , it have this function call \"kaggle_metric_utilities.safe_call_score(sklearn.metrics.roc_auc_score, solution[scored_columns].values, submission[scored_columns].values, average='macro'\", so could someone please clarify the same?
", "messages": [ { "id": "2822472", "postDate": "05/18/2024 15:55:31", "content": "I looked at the documentation of sklearn.metrics.roc_auc_score , i saw that there is an argument multi_class which needs to be passed as either 'ovo' or 'ovr' based on what kind of roc_auc we want for the case of multi class classfication otherwise it raises an error , as i checked in the metric given by kaggle , it have this function call \"kaggle_metric_utilities.safe_call_score(sklearn.metrics.roc_auc_score, solution[scored_columns].values, submission[scored_columns].values, average='macro'\", so could someone please clarify the same?
", "rawMarkdown": "I looked at the documentation of sklearn.metrics.roc_auc_score , i saw that there is an argument multi_class which needs to be passed as either 'ovo' or 'ovr' based on what kind of roc_auc we want for the case of multi class classfication otherwise it raises an error , as i checked in the metric given by kaggle , it have this function call \"**kaggle_metric_utilities.safe_call_score(sklearn.metrics.roc_auc_score, solution[scored_columns].values, submission[scored_columns].values, average='macro'**\", so could someone please clarify the same?", "votes": null }, { "id": "2822610", "postDate": "05/18/2024 17:19:28", "content": "I think those aren't used because it's supposed to be treated as a multilabel problem. It's unfortunate that the pinned notebooks are multiclass.
", "rawMarkdown": "I think those aren't used because it's supposed to be treated as a multilabel problem. It's unfortunate that the pinned notebooks are multiclass.", "votes": null }, { "id": "2822616", "postDate": "05/18/2024 17:20:48", "content": "The competition is a multi label competition. The metric is called with truth values as an array of size (n_samples, n_classes). In this case the parameter you cite is ignored. See the documentation: https://scikit-learn.org/stable/modules/generated/sklearn.metrics.roc_auc_score.html
", "rawMarkdown": "The competition is a multi label competition. The metric is called with truth values as an array of size (n_samples, n_classes). In this case the parameter you cite is ignored. See the documentation: https://scikit-learn.org/stable/modules/generated/sklearn.metrics.roc_auc_score.html", "votes": null }, { "id": "2822736", "postDate": "05/18/2024 18:25:29", "content": "It would be nice if you could use your powers as grand master to make the pinned notebooks be mutlilabel instead of multiclass.
", "rawMarkdown": "It would be nice if you could use your powers as grand master to make the pinned notebooks be mutlilabel instead of multiclass.", "votes": null }, { "id": "2822767", "postDate": "05/18/2024 18:39:55", "content": "I forgot how to read keras code I'm afraid.
", "rawMarkdown": "I forgot how to read keras code I'm afraid.", "votes": null } ], "comments": [ { "id": 2822610, "author_name": "willrice", "author_url": "", "post_date": "05/18/2024 17:19:28", "content": "I think those aren't used because it's supposed to be treated as a multilabel problem. It's unfortunate that the pinned notebooks are multiclass.
", "votes": null, "replies": [] }, { "id": 2822616, "author_name": "cpmpml", "author_url": "", "post_date": "05/18/2024 17:20:48", "content": "The competition is a multi label competition. The metric is called with truth values as an array of size (n_samples, n_classes). In this case the parameter you cite is ignored. See the documentation: https://scikit-learn.org/stable/modules/generated/sklearn.metrics.roc_auc_score.html
", "votes": null, "replies": [ { "id": 2822736, "author_name": "willrice", "author_url": "", "post_date": "05/18/2024 18:25:29", "content": "It would be nice if you could use your powers as grand master to make the pinned notebooks be mutlilabel instead of multiclass.
", "votes": null, "replies": [ { "id": 2822767, "author_name": "cpmpml", "author_url": "", "post_date": "05/18/2024 18:39:55", "content": "I forgot how to read keras code I'm afraid.
", "votes": null, "replies": [] } ] } ] } ], "raw_markdown_by_id": { "2822472": "I looked at the documentation of sklearn.metrics.roc_auc_score , i saw that there is an argument multi_class which needs to be passed as either 'ovo' or 'ovr' based on what kind of roc_auc we want for the case of multi class classfication otherwise it raises an error , as i checked in the metric given by kaggle , it have this function call \"**kaggle_metric_utilities.safe_call_score(sklearn.metrics.roc_auc_score, solution[scored_columns].values, submission[scored_columns].values, average='macro'**\", so could someone please clarify the same?", "2822610": "I think those aren't used because it's supposed to be treated as a multilabel problem. It's unfortunate that the pinned notebooks are multiclass.", "2822616": "The competition is a multi label competition. The metric is called with truth values as an array of size (n_samples, n_classes). In this case the parameter you cite is ignored. See the documentation: https://scikit-learn.org/stable/modules/generated/sklearn.metrics.roc_auc_score.html", "2822736": "It would be nice if you could use your powers as grand master to make the pinned notebooks be mutlilabel instead of multiclass.", "2822767": "I forgot how to read keras code I'm afraid." }, "source": "meta" }