{ "id": 94358, "title": "Useful lesson: simply generating lots of features isn't enough", "url": "/competitions/LANL-Earthquake-Prediction/discussion/94358", "author_name": "Andrey Lukyanenko", "post_date": "2019-06-04T04:18:07.511000", "votes": 12, "comment_count": 10, "views": 0, "content": "

Well, it was obvious in the very beginning, that generating hundreds of features won't guarantee the success (at the end I had ~8k features). I tried various approaches to feature selection, but still it wasn't enough :)

", "messages": [ { "id": 542739, "postDate": "2019-06-04T04:18:07.513Z", "content": "

Well, it was obvious in the very beginning, that generating hundreds of features won't guarantee the success (at the end I had ~8k features). I tried various approaches to feature selection, but still it wasn't enough :)

", "rawMarkdown": "Well, it was obvious in the very beginning, that generating hundreds of features won't guarantee the success (at the end I had ~8k features). I tried various approaches to feature selection, but still it wasn't enough :)", "votes": 11 }, { "id": 542836, "postDate": "2019-06-04T06:16:00.120Z", "content": "

I got a lot of insights from your kernel. Thank you.

", "rawMarkdown": "I got a lot of insights from your kernel. Thank you.", "votes": 5 }, { "id": 544094, "postDate": "2019-06-05T05:46:27.280Z", "content": "

Indeed, it was a very good lesson. I also computed way more features than needed, and I guess my methods to discard the bad ones fell short.

\n\n

I would also thank you for all your guidance. I really enjoyed going through your kernels.

", "rawMarkdown": "Indeed, it was a very good lesson. I also computed way more features than needed, and I guess my methods to discard the bad ones fell short.\n\nI would also thank you for all your guidance. I really enjoyed going through your kernels.", "votes": 1 }, { "id": 543104, "postDate": "2019-06-04T10:28:56.743Z", "content": "

Your content has been a great help. You gave everybody to try something. Thanks for sharing.

", "rawMarkdown": "Your content has been a great help. You gave everybody to try something. Thanks for sharing.", "votes": 2 }, { "id": 543093, "postDate": "2019-06-04T10:25:31.860Z", "content": "

Your FE inspired so many people here that you should get many warm thanks. I may have reused few of your features even, thanks you if you are the author!

", "rawMarkdown": "Your FE inspired so many people here that you should get many warm thanks. I may have reused few of your features even, thanks you if you are the author!", "votes": 2, "replies": [ { "id": 543242, "postDate": "2019-06-04T12:05:19.330Z", "content": "

Thank you! I know that most of these features weren't really useful, but even if several of them were important - it is great.

", "rawMarkdown": "Thank you! I know that most of these features weren't really useful, but even if several of them were important - it is great.", "votes": 1 } ] }, { "id": 542979, "postDate": "2019-06-04T08:49:55.427Z", "content": "

Thanks for sharing in this competition. I believe you inspired a lot of competitors like me.

", "rawMarkdown": "Thanks for sharing in this competition. I believe you inspired a lot of competitors like me. ", "votes": 2 }, { "id": 542833, "postDate": "2019-06-04T06:13:43.263Z", "content": "

You are still our hero though. One of our submission was using your features + XGB with a good hyper-parameter tuning using AWS achieved 2.4 in the private leaderboard.

", "rawMarkdown": "You are still our hero though. One of our submission was using your features + XGB with a good hyper-parameter tuning using AWS achieved 2.4 in the private leaderboard.", "votes": 2, "replies": [ { "id": 543238, "postDate": "2019-06-04T12:04:02.793Z", "content": "

Thank you :)\nWow, this is an impressive result!

", "rawMarkdown": "Thank you :)\nWow, this is an impressive result!" } ] }, { "id": 543111, "postDate": "2019-06-04T10:34:44.497Z", "rawMarkdown": "", "votes": 2, "isDeleted": true, "replies": [ { "id": 543241, "postDate": "2019-06-04T12:04:33.170Z", "content": "

Thank you! I'm really glad that my work helped so many people.

", "rawMarkdown": "Thank you! I'm really glad that my work helped so many people." } ] } ], "comments": [ { "id": 542836, "author_name": "Wonho Song", "author_url": "", "post_date": "2019-06-04T06:16:00.120000", "content": "

I got a lot of insights from your kernel. Thank you.

", "votes": 5, "replies": [] }, { "id": 544094, "author_name": "Ricard Delgado", "author_url": "", "post_date": "2019-06-05T05:46:27.280000", "content": "

Indeed, it was a very good lesson. I also computed way more features than needed, and I guess my methods to discard the bad ones fell short.

\n\n

I would also thank you for all your guidance. I really enjoyed going through your kernels.

", "votes": 1, "replies": [] }, { "id": 543104, "author_name": "Karan Jakhar", "author_url": "", "post_date": "2019-06-04T10:28:56.743000", "content": "

Your content has been a great help. You gave everybody to try something. Thanks for sharing.

", "votes": 2, "replies": [] }, { "id": 543093, "author_name": "CPMP", "author_url": "", "post_date": "2019-06-04T10:25:31.860000", "content": "

Your FE inspired so many people here that you should get many warm thanks. I may have reused few of your features even, thanks you if you are the author!

", "votes": 2, "replies": [ { "id": 543242, "author_name": "Andrey Lukyanenko", "author_url": "", "post_date": "2019-06-04T12:05:19.330000", "content": "

Thank you! I know that most of these features weren't really useful, but even if several of them were important - it is great.

", "votes": 1, "replies": [] } ] }, { "id": 542979, "author_name": "Morgon", "author_url": "", "post_date": "2019-06-04T08:49:55.427000", "content": "

Thanks for sharing in this competition. I believe you inspired a lot of competitors like me.

", "votes": 2, "replies": [] }, { "id": 542833, "author_name": "Shuhao Cao", "author_url": "", "post_date": "2019-06-04T06:13:43.263000", "content": "

You are still our hero though. One of our submission was using your features + XGB with a good hyper-parameter tuning using AWS achieved 2.4 in the private leaderboard.

", "votes": 2, "replies": [ { "id": 543238, "author_name": "Andrey Lukyanenko", "author_url": "", "post_date": "2019-06-04T12:04:02.793000", "content": "

Thank you :)\nWow, this is an impressive result!

", "votes": 0, "replies": [] } ] }, { "id": 543111, "author_name": "", "author_url": "", "post_date": "2019-06-04T10:34:44.497000", "content": "", "votes": 2, "replies": [ { "id": 543241, "author_name": "Andrey Lukyanenko", "author_url": "", "post_date": "2019-06-04T12:04:33.170000", "content": "

Thank you! I'm really glad that my work helped so many people.

", "votes": 0, "replies": [] } ] } ], "raw_markdown_by_id": { "542739": "Well, it was obvious in the very beginning, that generating hundreds of features won't guarantee the success (at the end I had ~8k features). I tried various approaches to feature selection, but still it wasn't enough :)", "542836": "I got a lot of insights from your kernel. Thank you.", "544094": "Indeed, it was a very good lesson. I also computed way more features than needed, and I guess my methods to discard the bad ones fell short.\n\nI would also thank you for all your guidance. I really enjoyed going through your kernels.", "543104": "Your content has been a great help. You gave everybody to try something. Thanks for sharing.", "543093": "Your FE inspired so many people here that you should get many warm thanks. I may have reused few of your features even, thanks you if you are the author!", "542979": "Thanks for sharing in this competition. I believe you inspired a lot of competitors like me. ", "542833": "You are still our hero though. One of our submission was using your features + XGB with a good hyper-parameter tuning using AWS achieved 2.4 in the private leaderboard.", "543111": "" } }