![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F8273630%2F83c8ae423e7414af0e8722f4e5b6ad2f%2FUnsel.png?generation=1734655494940392&alt=media\")
Hello all,
\nWhat was your best scoring submission that you did not select? 4-5 of my early submissions scored 0.464 here, that could have landed me at a rank of 31-35 perhaps!
\nWhat was your opportunity cost of an unselected submission here?
\n
Most of my early submissions scored well on the private leaderboard! Did you observe similarly?
", "messages": [ { "id": 3076411, "postDate": "2024-12-20T00:45:02.460Z", "content": "Hello all,
\nWhat was your best scoring submission that you did not select? 4-5 of my early submissions scored 0.464 here, that could have landed me at a rank of 31-35 perhaps!
\nWhat was your opportunity cost of an unselected submission here?
\n
Most of my early submissions scored well on the private leaderboard! Did you observe similarly?
", "rawMarkdown": "Hello all, \n\nWhat was your best scoring submission that you did not select? 4-5 of my early submissions scored 0.464 here, that could have landed me at a rank of 31-35 perhaps!\n\nWhat was your opportunity cost of an unselected submission here?\n\n\n\n# Key observation\nMost of my early submissions scored well on the private leaderboard! Did you observe similarly?", "votes": 13 }, { "id": 3076790, "postDate": "2024-12-20T09:13:42.617Z", "content": "![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F22136130%2Fec222f3c5da5ee5e72862a9c4672fa3f%2Fcmi.PNG?generation=1734685930629937&alt=media\")
đ
", "rawMarkdown": "\n\n[notebook](https://www.kaggle.com/code/tomyuen/clean-of-cmi-single-lgbm-f8db91?scriptVersionId=199839476)\n\nđ", "votes": 3, "replies": [ { "id": 3076826, "postDate": "2024-12-20T09:56:55.850Z", "content": "You could have been a prize winner here @tomyuen
\nIt will feel odd but this is the nature of such competitions
@tomyuen , sad to see , better luck next time.
", "rawMarkdown": "@tomyuen , sad to see , better luck next time. ", "votes": 1 } ] }, { "id": 3076419, "postDate": "2024-12-20T00:49:46.007Z", "content": "![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F13918144%2Fcc599b44e5366bcc3be8de89e58aa014%2F1734655679647.png?generation=1734655783433618&alt=media\")
I missed out on the prize money, but I'm thankful that I learned a lot from the competition
", "rawMarkdown": "I missed out on the prize money, but I'm thankful that I learned a lot from the competition", "votes": 3, "replies": [ { "id": 3076425, "postDate": "2024-12-20T00:52:43.850Z", "content": "Sad to see, the final standing is totally unexpected here. Mostly Model with Public LB 0.445 to 0.450 gives better private score.
", "rawMarkdown": "Sad to see, the final standing is totally unexpected here. Mostly Model with Public LB 0.445 to 0.450 gives better private score.", "votes": 1 }, { "id": 3076427, "postDate": "2024-12-20T00:53:47.690Z", "content": "Most laptops have been overfitting leaderboards since 0.45 by over-adjusting hyperparameters and so on
", "rawMarkdown": "Most laptops have been overfitting leaderboards since 0.45 by over-adjusting hyperparameters and so on", "votes": 1 } ] } ] }, { "id": 3076417, "postDate": "2024-12-20T00:48:45.273Z", "content": "Private Score : 0.459,
\nThis model perform worse as CV and Public Rank. But in Private it give a silver medal. But Missed
", "rawMarkdown": "Private Score : 0.459, \n\nThis model perform worse as CV and Public Rank. But in Private it give a silver medal. But Missed ", "votes": 3, "replies": [ { "id": 3076791, "postDate": "2024-12-20T09:14:21.317Z", "content": "I love your notebook @abdmental01. It's simply the best!
", "rawMarkdown": "I love your notebook @abdmental01. It's simply the best!", "votes": 2 } ] }, { "id": 3076990, "postDate": "2024-12-20T13:11:13.940Z", "content": "Here was my best unselected submission.
\nPrivate = 0.486, Public = 0.452
https://www.kaggle.com/code/motono0223/ros-seed42-cmi-new-target-sii-float?scriptVersionId=200323187
\nVersion 2 of 7
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F8163878%2F2ac4125ba84ceaa8dfba9617819c78e8%2FScreenshot%202024-12-20%2022.10.24.png?generation=1734700241421347&alt=media\")
[Main technique]
\nThis notebook created new object variable from the \"PCIAT_Total\" instead of \"sii\".
\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F8163878%2F9cbf9c36077037c2a35abace24fa550d%2FScreenshot%202024-12-20%2022.17.26.png?generation=1734700710577416&alt=media\")
\n(x axis: PCIAT_Total, y axis: objective values) (blue: sii, orange: new objective variable)
():\n pciat_total_value < :\n slope = \n offset = \n pciat_total_value < :\n slope = \n offset = -\n pciat_total_value < :\n slope = \n offset = \n :\n slope = \n offset = -\n slope * pciat_total_value + offset\nThis is beautiful! @motono0223
", "rawMarkdown": "This is beautiful! @motono0223 ", "votes": 1 }, { "id": 3077597, "postDate": "2024-12-21T06:48:31.377Z", "content": "Did your final submission include these techniques?
", "rawMarkdown": "Did your final submission include these techniques?", "votes": 1 }, { "id": 3077912, "postDate": "2024-12-21T14:11:40.533Z", "content": "@motono0223 how did you select the slope and offset? Is this a linear OLS model of some kind (looks like one to me)
", "rawMarkdown": "@motono0223 how did you select the slope and offset? Is this a linear OLS model of some kind (looks like one to me)" } ] }, { "id": 3076453, "postDate": "2024-12-20T01:09:49.797Z", "content": "I luckily chose the correct one of my highest public LB score notebooks for a silver medal.
\nBut my hatching choice was a miss :/
![\"choice\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F20325352%2F1f5a042a19e395bbb03713e18bfca58e%2Fwrong_one.PNG?generation=1734656894835652&alt=media\")
0.473 private
", "rawMarkdown": "0.473 private ", "votes": 1 }, { "id": 3077771, "postDate": "2024-12-21T11:38:31.100Z", "content": "Highest unselected was Private LB=0.460.
\nLink to notebook: https://www.kaggle.com/code/abhisekdash37/ensemble-learning-problematic-internet-use?scriptVersionId=205432552
Explanation: https://www.kaggle.com/competitions/child-mind-institute-problematic-internet-use/discussion/552753
", "rawMarkdown": "Highest unselected was Private LB=0.460. \nLink to notebook: https://www.kaggle.com/code/abhisekdash37/ensemble-learning-problematic-internet-use?scriptVersionId=205432552\n\nExplanation: https://www.kaggle.com/competitions/child-mind-institute-problematic-internet-use/discussion/552753", "votes": 1 }, { "id": 3077378, "postDate": "2024-12-20T21:32:19.317Z", "content": "My highest unselected was Private = 0.454, Public = 0.442
\nHere's how the LB scores varied with the versions -- there are two interesting jumps: i) Public LB jumped at v18, I had accidentally left out two important features: SDS-SDS_Total_T and PreInt_EduHx-computerinternet_hoursday; ii) Private LB jumped up at version 59 when I added train-quantile \"optimization\" of the y_hat values. After these a lot of meandering, ending with the Private consistantly below the PublicâŠ
All in all I enjoyed the addiction đ€Ș Code for the plot is all the way at the end of the CMI 2024 - Kaggle Addiction? notebook.
\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1496178%2F532d4e606eaf209e764142daa54654d3%2FPrivate_LB_vs_version.png?generation=1734729108980607&alt=media\")
Just changed n_splits to 65 from 5 in this public notebook:
\nhttps://www.kaggle.com/code/onodera/starter-notebook-with-polars-gpu
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F21354669%2F46dac257966d6346be4db449b5d024d3%2Ftemp.png?generation=1734672542728595&alt=media\")
You actually did 65 splits! Each split created an oof dataset of 40-43 rows! Is this right @ajayasaithambi
", "rawMarkdown": "You actually did 65 splits! Each split created an oof dataset of 40-43 rows! Is this right @ajayasaithambi " } ] }, { "id": 3076513, "postDate": "2024-12-20T02:38:33.210Z", "content": "My best score is 448. It is an ensemble of xgboost, tabnet, lightgbm and catboost, without finetuning anything.
", "rawMarkdown": "My best score is 448. It is an ensemble of xgboost, tabnet, lightgbm and catboost, without finetuning anything.", "votes": 1 }, { "id": 3076504, "postDate": "2024-12-20T02:23:01.223Z", "content": "My best unselected submission on private is 0.462(public 0.456)
\nThe best selected submission on private is 0.457(public 0.453)
I was surprised to get a silver.đ
\nI have found that high scoring notebooks have data leakage issues when using the KNN method. I was also misled at first, but luckily I found the problem and used a more stable method.
My score is 0.472!
\nAlthough I didnât make the final selection since the CV score wasnât exceptionally high, I realized the importance of analyzing the test items.
\nIn the notebooks I previously submitted with scores above 0.46, a theoretical approach to psychological test items proved to be effective.
\nFor example, conducting factor analysis to classify the 20 PCIAT items into four factors and using this information as features turned out to be effective in the private dataset.
\nHowever, this approach didnât show significant effects in the provided training data⊠sadly.
![\"Alt](\"https://i.imgur.com/Nj6g4bs.jpeg\")
licked gold but still ended with silver!
0.465! this was my early submission
", "rawMarkdown": "0.465! this was my early submission", "votes": 1 }, { "id": 3076429, "postDate": "2024-12-20T00:54:05.313Z", "content": "", "rawMarkdown": "PB:0.451", "votes": 1 }, { "id": 3076482, "postDate": "2024-12-20T01:49:04.750Z", "content": "A notebook without overfitting, CV:0.484,PB:0.481
\nMy notebook is a little confused, and specific predictions are In[37]
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F13918144%2Ffc4377464b194e286824657d7bc5e0e2%2F1734659156126.png?generation=1734659332684186&alt=media\")
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F13918144%2F242628b444ac109b8de83f5f22113947%2F2024122094825.bmp?generation=1734659399100828&alt=media\")
Best unselected is 0.475 [here].
\nAnd many between 0.460 to 4.70.
\nI am not kicking myself over it due to the fact there was in no way any indication that these were going to translate to private. If I had picked any of these solutions it was a matter of luck.
\nMy current silver solution that I did choose was the most stable and had the strongest indication (for me) to do well. It wasn't a top solution but happy with it nonetheless..
", "rawMarkdown": "Best unselected is 0.475 [[here](https://www.kaggle.com/code/bsmelbs/what-if)].\n\nAnd many between 0.460 to 4.70.\n\nI am not kicking myself over it due to the fact there was in no way any indication that these were going to translate to private. If I had picked any of these solutions it was a matter of luck.\n\nMy current silver solution that I did choose was the most stable and had the strongest indication (for me) to do well. It wasn't a top solution but happy with it nonetheless.. ", "votes": 2 }, { "id": 3076423, "postDate": "2024-12-20T00:51:23.030Z", "content": "0.455 , A solution with both low CV and LBđ
", "rawMarkdown": "0.455 , A solution with both low CV and LBđ\n", "votes": 2, "replies": [ { "id": 3076428, "postDate": "2024-12-20T00:53:57.077Z", "content": "all the good scores are lie between 0.445 to 0.450 on public LB.
", "rawMarkdown": "all the good scores are lie between 0.445 to 0.450 on public LB.", "votes": 3, "replies": [ { "id": 3078762, "postDate": "2024-12-22T19:12:55.303Z", "content": "I got 0.47 on private and 0.469 on public
", "rawMarkdown": "I got 0.47 on private and 0.469 on public", "votes": 1 } ] } ] }, { "id": 3076414, "postDate": "2024-12-20T00:47:13.757Z", "content": "Iâm waiting to see someone with a notebook that has a private score of 0.50 but wasnât selected. Letâs waitâitâs only a matter of time, and weâll see them in the next few hours! âłđ
", "rawMarkdown": "Iâm waiting to see someone with a notebook that has a private score of 0.50 but wasnât selected. Letâs waitâitâs only a matter of time, and weâll see them in the next few hours! âłđ", "votes": 2, "replies": [ { "id": 3080150, "postDate": "2024-12-24T18:10:33.963Z", "rawMarkdown": "", "votes": 1, "isDeleted": true } ] }, { "id": 3185957, "postDate": "2025-04-24T03:43:53.617Z", "content": "Hi Ravy, Do you still remember this competition? When I was organizing my warehouse today, I found a notebook with a very high score? He got a score of 0.497 on the private test set. Is this the highest score on the private test set?
", "rawMarkdown": "Hi Ravy, Do you still remember this competition? When I was organizing my warehouse today, I found a notebook with a very high score? He got a score of 0.497 on the private test set. Is this the highest score on the private test set? " }, { "id": 3077942, "postDate": "2024-12-21T15:18:11.997Z", "content": "Here was my best unselected submission.
\nPrivate = 0.486, Public = 0.452
đ
", "votes": 3, "replies": [ { "id": 3076826, "author_name": "Ravi Ramakrishnan", "author_url": "", "post_date": "2024-12-20T09:56:55.850000", "content": "You could have been a prize winner here @tomyuen
\nIt will feel odd but this is the nature of such competitions
@tomyuen , sad to see , better luck next time.
", "votes": 1, "replies": [] } ] }, { "id": 3076419, "author_name": "Reki", "author_url": "", "post_date": "2024-12-20T00:49:46.007000", "content": "
I missed out on the prize money, but I'm thankful that I learned a lot from the competition
", "votes": 3, "replies": [ { "id": 3076425, "author_name": "Sheikh Muhammad Abdullah", "author_url": "", "post_date": "2024-12-20T00:52:43.850000", "content": "Sad to see, the final standing is totally unexpected here. Mostly Model with Public LB 0.445 to 0.450 gives better private score.
", "votes": 1, "replies": [] }, { "id": 3076427, "author_name": "Reki", "author_url": "", "post_date": "2024-12-20T00:53:47.690000", "content": "Most laptops have been overfitting leaderboards since 0.45 by over-adjusting hyperparameters and so on
", "votes": 1, "replies": [] } ] } ] }, { "id": 3076417, "author_name": "Sheikh Muhammad Abdullah", "author_url": "", "post_date": "2024-12-20T00:48:45.273000", "content": "Private Score : 0.459,
\nThis model perform worse as CV and Public Rank. But in Private it give a silver medal. But Missed
", "votes": 3, "replies": [ { "id": 3076791, "author_name": "Tom Yuen", "author_url": "", "post_date": "2024-12-20T09:14:21.317000", "content": "I love your notebook @abdmental01. It's simply the best!
", "votes": 2, "replies": [] } ] }, { "id": 3076990, "author_name": "motono0223", "author_url": "", "post_date": "2024-12-20T13:11:13.940000", "content": "Here was my best unselected submission.
\nPrivate = 0.486, Public = 0.452
https://www.kaggle.com/code/motono0223/ros-seed42-cmi-new-target-sii-float?scriptVersionId=200323187
\nVersion 2 of 7
[Main technique]
\nThis notebook created new object variable from the \"PCIAT_Total\" instead of \"sii\".
\n
\n(x axis: PCIAT_Total, y axis: objective values) (blue: sii, orange: new objective variable)
():\n pciat_total_value < :\n slope = \n offset = \n pciat_total_value < :\n slope = \n offset = -\n pciat_total_value < :\n slope = \n offset = \n :\n slope = \n offset = -\n slope * pciat_total_value + offset\nThis is beautiful! @motono0223
", "votes": 1, "replies": [] }, { "id": 3077597, "author_name": "Ajay Asaithambi", "author_url": "", "post_date": "2024-12-21T06:48:31.377000", "content": "Did your final submission include these techniques?
", "votes": 1, "replies": [] }, { "id": 3077912, "author_name": "Ravi Ramakrishnan", "author_url": "", "post_date": "2024-12-21T14:11:40.533000", "content": "@motono0223 how did you select the slope and offset? Is this a linear OLS model of some kind (looks like one to me)
", "votes": 0, "replies": [] } ] }, { "id": 3076453, "author_name": "Marius Heuser", "author_url": "", "post_date": "2024-12-20T01:09:49.797000", "content": "I luckily chose the correct one of my highest public LB score notebooks for a silver medal.
\nBut my hatching choice was a miss :/
0.473 private
", "votes": 1, "replies": [] }, { "id": 3077771, "author_name": "Abhisek Dash", "author_url": "", "post_date": "2024-12-21T11:38:31.100000", "content": "Highest unselected was Private LB=0.460.
\nLink to notebook: https://www.kaggle.com/code/abhisekdash37/ensemble-learning-problematic-internet-use?scriptVersionId=205432552
Explanation: https://www.kaggle.com/competitions/child-mind-institute-problematic-internet-use/discussion/552753
", "votes": 1, "replies": [] }, { "id": 3077378, "author_name": "Daniel Dewey", "author_url": "", "post_date": "2024-12-20T21:32:19.317000", "content": "My highest unselected was Private = 0.454, Public = 0.442
\nHere's how the LB scores varied with the versions -- there are two interesting jumps: i) Public LB jumped at v18, I had accidentally left out two important features: SDS-SDS_Total_T and PreInt_EduHx-computerinternet_hoursday; ii) Private LB jumped up at version 59 when I added train-quantile \"optimization\" of the y_hat values. After these a lot of meandering, ending with the Private consistantly below the PublicâŠ
All in all I enjoyed the addiction đ€Ș Code for the plot is all the way at the end of the CMI 2024 - Kaggle Addiction? notebook.
\n
Just changed n_splits to 65 from 5 in this public notebook:
\nhttps://www.kaggle.com/code/onodera/starter-notebook-with-polars-gpu
You actually did 65 splits! Each split created an oof dataset of 40-43 rows! Is this right @ajayasaithambi
", "votes": 0, "replies": [] } ] }, { "id": 3076513, "author_name": "Anh Bui", "author_url": "", "post_date": "2024-12-20T02:38:33.210000", "content": "My best score is 448. It is an ensemble of xgboost, tabnet, lightgbm and catboost, without finetuning anything.
", "votes": 1, "replies": [] }, { "id": 3076504, "author_name": "SnakeWoodMan", "author_url": "", "post_date": "2024-12-20T02:23:01.223000", "content": "My best unselected submission on private is 0.462(public 0.456)
\nThe best selected submission on private is 0.457(public 0.453)
I was surprised to get a silver.đ
\nI have found that high scoring notebooks have data leakage issues when using the KNN method. I was also misled at first, but luckily I found the problem and used a more stable method.
My score is 0.472!
\nAlthough I didnât make the final selection since the CV score wasnât exceptionally high, I realized the importance of analyzing the test items.
\nIn the notebooks I previously submitted with scores above 0.46, a theoretical approach to psychological test items proved to be effective.
\nFor example, conducting factor analysis to classify the 20 PCIAT items into four factors and using this information as features turned out to be effective in the private dataset.
\nHowever, this approach didnât show significant effects in the provided training data⊠sadly.
licked gold but still ended with silver!
0.465! this was my early submission
", "votes": 1, "replies": [] }, { "id": 3076429, "author_name": "yunsuxiaozi", "author_url": "", "post_date": "2024-12-20T00:54:05.313000", "content": "", "votes": 1, "replies": [] }, { "id": 3076482, "author_name": "Reki", "author_url": "", "post_date": "2024-12-20T01:49:04.750000", "content": "A notebook without overfitting, CV:0.484,PB:0.481
\nMy notebook is a little confused, and specific predictions are In[37]
Best unselected is 0.475 [here].
\nAnd many between 0.460 to 4.70.
\nI am not kicking myself over it due to the fact there was in no way any indication that these were going to translate to private. If I had picked any of these solutions it was a matter of luck.
\nMy current silver solution that I did choose was the most stable and had the strongest indication (for me) to do well. It wasn't a top solution but happy with it nonetheless..
", "votes": 2, "replies": [] }, { "id": 3076423, "author_name": "passionfruit216", "author_url": "", "post_date": "2024-12-20T00:51:23.030000", "content": "0.455 , A solution with both low CV and LBđ
", "votes": 2, "replies": [ { "id": 3076428, "author_name": "Sheikh Muhammad Abdullah", "author_url": "", "post_date": "2024-12-20T00:53:57.077000", "content": "all the good scores are lie between 0.445 to 0.450 on public LB.
", "votes": 3, "replies": [ { "id": 3078762, "author_name": "RafaĆ PawĆowski", "author_url": "", "post_date": "2024-12-22T19:12:55.303000", "content": "I got 0.47 on private and 0.469 on public
", "votes": 1, "replies": [] } ] } ] }, { "id": 3076414, "author_name": "Rabie El Kharoua", "author_url": "", "post_date": "2024-12-20T00:47:13.757000", "content": "Iâm waiting to see someone with a notebook that has a private score of 0.50 but wasnât selected. Letâs waitâitâs only a matter of time, and weâll see them in the next few hours! âłđ
", "votes": 2, "replies": [ { "id": 3080150, "author_name": "", "author_url": "", "post_date": "2024-12-24T18:10:33.963000", "content": "", "votes": 1, "replies": [] } ] }, { "id": 3185957, "author_name": "Reki", "author_url": "", "post_date": "2025-04-24T03:43:53.617000", "content": "Hi Ravy, Do you still remember this competition? When I was organizing my warehouse today, I found a notebook with a very high score? He got a score of 0.497 on the private test set. Is this the highest score on the private test set?
", "votes": 0, "replies": [] }, { "id": 3077942, "author_name": "MrSimple", "author_url": "", "post_date": "2024-12-21T15:18:11.997000", "content": "Here was my best unselected submission.
\nPrivate = 0.486, Public = 0.452