I am a newbie in data science. Struggle for a week, Luckily, I get bronze temporarily.\nBut, my CV and LB get stuck at 1.9798 and 1.424. The next work is to reduce features, but I have not any experience with statistics. So, I can only use selectKbest() to drop some features, but it is useless for my model. \nIf I drop too many features, LB gets really bad.\nIf it's convenient, please tell me some examples, tips or functions abou it.
", "messages": [ { "id": 532619, "postDate": "2019-05-17T11:40:08.497Z", "content": "you may need this: https://www.researchgate.net/publication/221996079_An_Introduction_of_Variable_and_Feature_Selection
", "rawMarkdown": "you may need this: https://www.researchgate.net/publication/221996079_An_Introduction_of_Variable_and_Feature_Selection", "votes": 5, "replies": [ { "id": 532621, "postDate": "2019-05-17T11:45:45.837Z", "content": "谢谢啦~
", "rawMarkdown": "谢谢啦~", "votes": -2 }, { "id": 537865, "postDate": "2019-05-27T19:07:23.207Z", "content": "Thank you for this link! It is helpful.
", "rawMarkdown": "Thank you for this link! It is helpful." } ] }, { "id": 532605, "postDate": "2019-05-17T11:13:24.547Z", "content": "try feature importance in lgb, or pearsonr (from scipy.stats import pearsonr) with different thresholds (look at vettejeep kernel)
", "rawMarkdown": "try feature importance in lgb, or pearsonr (from scipy.stats import pearsonr) with different thresholds (look at vettejeep kernel) \n", "votes": 5, "replies": [ { "id": 532622, "postDate": "2019-05-17T11:46:14.503Z", "content": "Thanks a lot
", "rawMarkdown": "Thanks a lot" } ] }, { "id": 533647, "postDate": "2019-05-19T17:16:51.073Z", "content": "I plot pictures, I shared one way. You can also plot partial dependency graph.
", "rawMarkdown": "I plot pictures, I shared one way. You can also plot partial dependency graph.", "votes": 3 }, { "id": 532511, "postDate": "2019-05-17T06:27:30.387Z", "content": "I am a newbie in data science. Struggle for a week, Luckily, I get bronze temporarily.\nBut, my CV and LB get stuck at 1.9798 and 1.424. The next work is to reduce features, but I have not any experience with statistics. So, I can only use selectKbest() to drop some features, but it is useless for my model. \nIf I drop too many features, LB gets really bad.\nIf it's convenient, please tell me some examples, tips or functions abou it.
", "rawMarkdown": "I am a newbie in data science. Struggle for a week, Luckily, I get bronze temporarily.\nBut, my CV and LB get stuck at 1.9798 and 1.424. The next work is to reduce features, but I have not any experience with statistics. So, I can only use selectKbest() to drop some features, but it is useless for my model. \nIf I drop too many features, LB gets really bad.\nIf it's convenient, please tell me some examples, tips or functions abou it.\n\n", "votes": 2 }, { "id": 539137, "postDate": "2019-05-29T15:23:00.107Z", "content": "@haonanyao look at this: https://towardsdatascience.com/feature-selection-techniques-in-machine-learning-with-python-f24e7da3f36e
", "rawMarkdown": "@haonanyao look at this: https://towardsdatascience.com/feature-selection-techniques-in-machine-learning-with-python-f24e7da3f36e" }, { "id": 533096, "postDate": "2019-05-18T12:58:46.213Z", "content": "Try chi2 or some other feature selection technique
", "rawMarkdown": "Try chi2 or some other feature selection technique " }, { "id": 532907, "postDate": "2019-05-18T02:32:31.057Z", "rawMarkdown": "", "isDeleted": true, "replies": [ { "id": 533375, "postDate": "2019-05-19T05:12:25.003Z", "content": "200 features CV-LB:1.92-1.440\nbut my best LB have 2000 featues.
", "rawMarkdown": "200 features CV-LB:1.92-1.440\nbut my best LB have 2000 featues.", "votes": 1 } ] } ], "comments": [ { "id": 532619, "author_name": "MeisterMorxrc", "author_url": "", "post_date": "2019-05-17T11:40:08.497000", "content": "you may need this: https://www.researchgate.net/publication/221996079_An_Introduction_of_Variable_and_Feature_Selection
", "votes": 5, "replies": [ { "id": 532621, "author_name": "Haonan Yao", "author_url": "", "post_date": "2019-05-17T11:45:45.837000", "content": "谢谢啦~
", "votes": -2, "replies": [] }, { "id": 537865, "author_name": "DevilEars", "author_url": "", "post_date": "2019-05-27T19:07:23.207000", "content": "Thank you for this link! It is helpful.
", "votes": 0, "replies": [] } ] }, { "id": 532605, "author_name": "Matias Thayer", "author_url": "", "post_date": "2019-05-17T11:13:24.547000", "content": "try feature importance in lgb, or pearsonr (from scipy.stats import pearsonr) with different thresholds (look at vettejeep kernel)
", "votes": 5, "replies": [ { "id": 532622, "author_name": "Haonan Yao", "author_url": "", "post_date": "2019-05-17T11:46:14.503000", "content": "Thanks a lot
", "votes": 0, "replies": [] } ] }, { "id": 533647, "author_name": "CPMP", "author_url": "", "post_date": "2019-05-19T17:16:51.073000", "content": "I plot pictures, I shared one way. You can also plot partial dependency graph.
", "votes": 3, "replies": [] }, { "id": 539137, "author_name": "Felipe Loque", "author_url": "", "post_date": "2019-05-29T15:23:00.107000", "content": "@haonanyao look at this: https://towardsdatascience.com/feature-selection-techniques-in-machine-learning-with-python-f24e7da3f36e
", "votes": 0, "replies": [] }, { "id": 533096, "author_name": "prafful baheti", "author_url": "", "post_date": "2019-05-18T12:58:46.213000", "content": "Try chi2 or some other feature selection technique
", "votes": 0, "replies": [] }, { "id": 532907, "author_name": "", "author_url": "", "post_date": "2019-05-18T02:32:31.057000", "content": "", "votes": 0, "replies": [ { "id": 533375, "author_name": "Haonan Yao", "author_url": "", "post_date": "2019-05-19T05:12:25.003000", "content": "200 features CV-LB:1.92-1.440\nbut my best LB have 2000 featues.
", "votes": 1, "replies": [] } ] } ], "raw_markdown_by_id": { "532619": "you may need this: https://www.researchgate.net/publication/221996079_An_Introduction_of_Variable_and_Feature_Selection", "532605": "try feature importance in lgb, or pearsonr (from scipy.stats import pearsonr) with different thresholds (look at vettejeep kernel) \n", "533647": "I plot pictures, I shared one way. You can also plot partial dependency graph.", "532511": "I am a newbie in data science. Struggle for a week, Luckily, I get bronze temporarily.\nBut, my CV and LB get stuck at 1.9798 and 1.424. The next work is to reduce features, but I have not any experience with statistics. So, I can only use selectKbest() to drop some features, but it is useless for my model. \nIf I drop too many features, LB gets really bad.\nIf it's convenient, please tell me some examples, tips or functions abou it.\n\n", "539137": "@haonanyao look at this: https://towardsdatascience.com/feature-selection-techniques-in-machine-learning-with-python-f24e7da3f36e", "533096": "Try chi2 or some other feature selection technique ", "532907": "" } }