I have reviewed the best solutions of winners (gold medalists) of the NCAA-competitions in previous 2018-2019 (Men's and Women's) and provided links to their kernels and posts in section \"1.10. Google Cloud & NCAA® ML Competitions\" of the kernel Data Science for tabular data: Advanced Techniques (80 upvotes).
\n\nInterestingly, that:\n- organizers of the competition changes the rules and conditions from year to year, so the kernels need to be significantly improved (so I didn't analyze the competition before 2017);
\n\nthe most popular models of winners: linear regression and xgboost;
some have used models from @raddar in different years, but there have been some own developments;
most winners used R, though some also used Python (some developed a solution on R, and later adapted it to Python for promotion);
the authors write that their best decisions in these competitions for men are bad for women and vice versa, since each competition requires adaptation to their specificity;
the best kernels used given risky assumptions about winning certain basketball teams (the risk turned out to be justified, and one winner came in second place, thanks to the very risky prediction he made without knowing the capabilities of the teams - approached tasks as an ML specialist, and not basketball player).
I hope you find this kernel useful and enjoyable.\nYour comments and feedback are most welcome.
", "messages": [ { "id": "766834", "postDate": "03/08/2020 20:11:30", "content": "I have reviewed the best solutions of winners (gold medalists) of the NCAA-competitions in previous 2018-2019 (Men's and Women's) and provided links to their kernels and posts in section \"1.10. Google Cloud & NCAA® ML Competitions\" of the kernel Data Science for tabular data: Advanced Techniques (80 upvotes).
\n\nInterestingly, that:\n- organizers of the competition changes the rules and conditions from year to year, so the kernels need to be significantly improved (so I didn't analyze the competition before 2017);
\n\nthe most popular models of winners: linear regression and xgboost;
some have used models from @raddar in different years, but there have been some own developments;
most winners used R, though some also used Python (some developed a solution on R, and later adapted it to Python for promotion);
the authors write that their best decisions in these competitions for men are bad for women and vice versa, since each competition requires adaptation to their specificity;
the best kernels used given risky assumptions about winning certain basketball teams (the risk turned out to be justified, and one winner came in second place, thanks to the very risky prediction he made without knowing the capabilities of the teams - approached tasks as an ML specialist, and not basketball player).
I hope you find this kernel useful and enjoyable.\nYour comments and feedback are most welcome.
", "rawMarkdown": "I have reviewed the best solutions of winners (gold medalists) of the NCAA-competitions in previous 2018-2019 (Men's and Women's) and provided links to their kernels and posts in section \"1.10. Google Cloud & NCAA® ML Competitions\" of the kernel [Data Science for tabular data: Advanced Techniques](https://www.kaggle.com/vbmokin/data-science-for-tabular-data-advanced-techniques) (80 upvotes). \n\nInterestingly, that:\n- organizers of the competition changes the rules and conditions from year to year, so the kernels need to be significantly improved (so I didn't analyze the competition before 2017);\n\n- the most popular models of winners: linear regression and xgboost;\n\n- some have used models from @raddar in different years, but there have been some own developments;\n\n- most winners used R, though some also used Python (some developed a solution on R, and later adapted it to Python for promotion);\n\n- the authors write that their best decisions in these competitions for men are bad for women and vice versa, since each competition requires adaptation to their specificity;\n\n- the best kernels used given risky assumptions about winning certain basketball teams (the risk turned out to be justified, and one winner came in second place, thanks to the very risky prediction he made without knowing the capabilities of the teams - approached tasks as an ML specialist, and not basketball player).\n\nI hope you find this kernel useful and enjoyable.\nYour comments and feedback are most welcome.", "votes": null }, { "id": "766974", "postDate": "03/09/2020 02:02:11", "content": "The decision to use R or Python has minimal impact. I think you'll see more Python - nearly exclusive - as it's more popular than R. Last year I wrote everything in SQL and R. This year it'll be only Python.
\nI have found that the Men's and Women's comps are very different. I've made significant changes to the approach for each. Last year I didn't and I medaled in the Women's but finished >500th in the Men's.
Yes, Python became more popular than R. I also wrote earlier on R, and in recent years switched to Python. But it is useful to know both languages. There are many good kernels with interesting solutions in Kaggle in both languages.
", "rawMarkdown": "Yes, Python became more popular than R. I also wrote earlier on R, and in recent years switched to Python. But it is useful to know both languages. There are many good kernels with interesting solutions in Kaggle in both languages.", "votes": null }, { "id": "767853", "postDate": "03/10/2020 07:03:28", "content": "@vbmokin It seems like the link is not working. Could you please verify that?
\n\nGreat insight by the way. :)
", "rawMarkdown": "vbmokin It seems like the link is not working. Could you please verify that?\n\nGreat insight by the way. :)", "votes": null }, { "id": "768205", "postDate": "03/10/2020 14:40:36", "content": "I fixed it, thanks for your message
", "rawMarkdown": "I fixed it, thanks for your message", "votes": null } ], "comments": [ { "id": 766974, "author_name": "btesch", "author_url": "", "post_date": "03/09/2020 02:02:11", "content": "The decision to use R or Python has minimal impact. I think you'll see more Python - nearly exclusive - as it's more popular than R. Last year I wrote everything in SQL and R. This year it'll be only Python.
\nI have found that the Men's and Women's comps are very different. I've made significant changes to the approach for each. Last year I didn't and I medaled in the Women's but finished >500th in the Men's.
Yes, Python became more popular than R. I also wrote earlier on R, and in recent years switched to Python. But it is useful to know both languages. There are many good kernels with interesting solutions in Kaggle in both languages.
", "votes": null, "replies": [] } ] }, { "id": 767853, "author_name": "samuelkurabachew", "author_url": "", "post_date": "03/10/2020 07:03:28", "content": "@vbmokin It seems like the link is not working. Could you please verify that?
\n\nGreat insight by the way. :)
", "votes": null, "replies": [ { "id": 768205, "author_name": "vbmokin", "author_url": "", "post_date": "03/10/2020 14:40:36", "content": "I fixed it, thanks for your message
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "766834": "I have reviewed the best solutions of winners (gold medalists) of the NCAA-competitions in previous 2018-2019 (Men's and Women's) and provided links to their kernels and posts in section \"1.10. Google Cloud & NCAA® ML Competitions\" of the kernel [Data Science for tabular data: Advanced Techniques](https://www.kaggle.com/vbmokin/data-science-for-tabular-data-advanced-techniques) (80 upvotes). \n\nInterestingly, that:\n- organizers of the competition changes the rules and conditions from year to year, so the kernels need to be significantly improved (so I didn't analyze the competition before 2017);\n\n- the most popular models of winners: linear regression and xgboost;\n\n- some have used models from @raddar in different years, but there have been some own developments;\n\n- most winners used R, though some also used Python (some developed a solution on R, and later adapted it to Python for promotion);\n\n- the authors write that their best decisions in these competitions for men are bad for women and vice versa, since each competition requires adaptation to their specificity;\n\n- the best kernels used given risky assumptions about winning certain basketball teams (the risk turned out to be justified, and one winner came in second place, thanks to the very risky prediction he made without knowing the capabilities of the teams - approached tasks as an ML specialist, and not basketball player).\n\nI hope you find this kernel useful and enjoyable.\nYour comments and feedback are most welcome.", "766974": "The decision to use R or Python has minimal impact. I think you'll see more Python - nearly exclusive - as it's more popular than R. Last year I wrote everything in SQL and R. This year it'll be only Python. \nI have found that the Men's and Women's comps are very different. I've made significant changes to the approach for each. Last year I didn't and I medaled in the Women's but finished >500th in the Men's.", "767411": "Yes, Python became more popular than R. I also wrote earlier on R, and in recent years switched to Python. But it is useful to know both languages. There are many good kernels with interesting solutions in Kaggle in both languages.", "767853": "vbmokin It seems like the link is not working. Could you please verify that?\n\nGreat insight by the way. :)", "768205": "I fixed it, thanks for your message" }, "source": "meta" }