![](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F8273630%2Fc9176108d45b97dd673ac2d5c60d3583%2FKaggleComps.png?generation=1710233742687274&alt=media\")
Hey thanks for making time to read this.
\nI was able to get started thank you to the helpful starter notebook @jetakow built.
\nFor those of you with experience in multiple PA competitions, what is typically your first step to improving a model?
\nI am curious for both the competition and my personal development because it really has me stumped.
\nTotally get you might be busy so I won't expect you to respond, but any feedback would mean a lot.
\nCheers,
\nEmin Okic
Hey thanks for making time to read this.
\nI was able to get started thank you to the helpful starter notebook @jetakow built.
\nFor those of you with experience in multiple PA competitions, what is typically your first step to improving a model?
\nI am curious for both the competition and my personal development because it really has me stumped.
\nTotally get you might be busy so I won't expect you to respond, but any feedback would mean a lot.
\nCheers,
\nEmin Okic
Feature engineering is a priority. Take time with this and don't rush to submit subpar models. You may invest a week for your initial FE followed by a meaningful submission and then move ahead iteratively. This should continue till the end of the challenge, but you will be a better position to submit frequently after your code pipeline is ready @eminokic
\nPlease peruse the picture below for details
\n
I am glad you found that notebook helpful. When we develop models internally the FE part is the most important part of the whole model development. Besides the usual techniques you should think of a way how to make your model robust enough for changing the underlying structure of data in future.
", "rawMarkdown": "I am glad you found that notebook helpful. When we develop models internally the FE part is the most important part of the whole model development. Besides the usual techniques you should think of a way how to make your model robust enough for changing the underlying structure of data in future.", "votes": null }, { "id": "2693775", "postDate": "03/12/2024 16:44:34", "content": "Thanks Daniel. It means a lot that you made time to respond
", "rawMarkdown": "Thanks Daniel. It means a lot that you made time to respond", "votes": null }, { "id": "2693777", "postDate": "03/12/2024 16:44:49", "content": "Thanks Ravi! This is awesome. I appreciate you
", "rawMarkdown": "Thanks Ravi! This is awesome. I appreciate you", "votes": null }, { "id": "2693838", "postDate": "03/12/2024 17:49:09", "content": "No problem, I try to reply to every meaningful question, but it seems like I can't keep up with everything here :)
", "rawMarkdown": "No problem, I try to reply to every meaningful question, but it seems like I can't keep up with everything here :)", "votes": null } ], "comments": [ { "id": 2693148, "author_name": "ravi20076", "author_url": "", "post_date": "03/12/2024 08:49:01", "content": "Feature engineering is a priority. Take time with this and don't rush to submit subpar models. You may invest a week for your initial FE followed by a meaningful submission and then move ahead iteratively. This should continue till the end of the challenge, but you will be a better position to submit frequently after your code pipeline is ready @eminokic
\nPlease peruse the picture below for details
\n
Thanks Ravi! This is awesome. I appreciate you
", "votes": null, "replies": [] } ] }, { "id": 2693359, "author_name": "jetakow", "author_url": "", "post_date": "03/12/2024 11:38:30", "content": "I am glad you found that notebook helpful. When we develop models internally the FE part is the most important part of the whole model development. Besides the usual techniques you should think of a way how to make your model robust enough for changing the underlying structure of data in future.
", "votes": null, "replies": [ { "id": 2693775, "author_name": "eminokic", "author_url": "", "post_date": "03/12/2024 16:44:34", "content": "Thanks Daniel. It means a lot that you made time to respond
", "votes": null, "replies": [ { "id": 2693838, "author_name": "jetakow", "author_url": "", "post_date": "03/12/2024 17:49:09", "content": "No problem, I try to reply to every meaningful question, but it seems like I can't keep up with everything here :)
", "votes": null, "replies": [] } ] } ] } ], "raw_markdown_by_id": { "2692515": "Hey thanks for making time to read this. \n\nI was able to get started thank you to the helpful starter notebook @jetakow built. \n\nFor those of you with experience in multiple PA competitions, what is typically your first step to improving a model?\n\nI am curious for both the competition and my personal development because it really has me stumped. \n\nTotally get you might be busy so I won't expect you to respond, but any feedback would mean a lot. \n\nCheers,\nEmin Okic", "2693148": "Feature engineering is a priority. Take time with this and don't rush to submit subpar models. You may invest a week for your initial FE followed by a meaningful submission and then move ahead iteratively. This should continue till the end of the challenge, but you will be a better position to submit frequently after your code pipeline is ready @eminokic \n\nPlease peruse the picture below for details\n\n", "2693359": "I am glad you found that notebook helpful. When we develop models internally the FE part is the most important part of the whole model development. Besides the usual techniques you should think of a way how to make your model robust enough for changing the underlying structure of data in future.", "2693775": "Thanks Daniel. It means a lot that you made time to respond", "2693777": "Thanks Ravi! This is awesome. I appreciate you", "2693838": "No problem, I try to reply to every meaningful question, but it seems like I can't keep up with everything here :)" }, "source": "meta" }