Hello everyone! I am very happy to have won a gold medal and I would like to share my solution with you:
\nMany thanks to the organizers of this competition and to everyone who participated by sharing their ideas and code !
\nHere I share the code of my solution.
", "messages": [ { "id": "3076558", "postDate": "12/20/2024 04:22:56", "content": "Hello everyone! I am very happy to have won a gold medal and I would like to share my solution with you:
\nMany thanks to the organizers of this competition and to everyone who participated by sharing their ideas and code !
\nHere I share the code of my solution.
", "rawMarkdown": "Hello everyone! I am very happy to have won a gold medal and I would like to share my solution with you:\n- This is a single classification model (yes! not a regression model)\n- This solution was based on some sophisticated DNN that was the winning solution of the ICR competition. I loved it when I saw it and I was waiting for a competition where I could apply it.\n- Use IterativeImputer class as a strategy for imputing the missing values, which models each feature with missing values based on other features and uses that estimate for imputation. I decided to use both the train and test data to perform the fit and trained the model online so that I could take the data from the private test.\n- Take the enmo from the actigraphy data differentiating weekdays from weekends and grouping them by day and time slots.\n- Train with ALL the data\n- Use weights to balance the classes in the loss function\n- Take the average of the predictions of the same model with 10 different seeds\n\nMany thanks to the organizers of this competition and to everyone who participated by sharing their ideas and code !\n\nHere I share the code of my solution.", "votes": null }, { "id": "3076563", "postDate": "12/20/2024 04:30:48", "content": "Congratulations. I was also inspired by this winning DNN but used another one. IterativeImputer produced good results in my pipeline as well; I preferred this technique among other imputation ways.
Felicitaciones Laura! Increíble shakeup, gran trabajo. Buen autoregalo de navidad la medalla de oro ja. Felices fiestas!
", "rawMarkdown": "Felicitaciones Laura! Increíble shakeup, gran trabajo. Buen autoregalo de navidad la medalla de oro ja. Felices fiestas!", "votes": null }, { "id": "3077263", "postDate": "12/20/2024 18:07:00", "content": "Muchas gracias Maxi ! Muy contenta con mi autoregalito Navideño, jaja. Felices fiestas !
", "rawMarkdown": "Muchas gracias Maxi ! Muy contenta con mi autoregalito Navideño, jaja. Felices fiestas !", "votes": null }, { "id": "3078698", "postDate": "12/22/2024 17:23:58", "content": "@lauraromar firstly congratulation , isee that you participate in ICR competition you are not get good position in it but it seems that you learn a lot from it, what i learn from your post is that when you work hard in a competition you never lose , if you get gold or silver medal that good and if you get nothing you'll be prepared fr next competitions what is important is to keep going
", "rawMarkdown": "lauraromar firstly congratulation , isee that you participate in ICR competition you are not get good position in it but it seems that you learn a lot from it, what i learn from your post is that when you work hard in a competition you never lose , if you get gold or silver medal that good and if you get nothing you'll be prepared fr next competitions what is important is to keep going", "votes": null }, { "id": "3078882", "postDate": "12/23/2024 01:21:47", "content": "Thank you so much ! It's just as you say, in my case, I learned a lot from each competition, thanks to the invaluable discussions and the code shared by the kagglers. I think this is the most valuable thing that competitions leave us…beyond the medals that always make us happy :)
", "rawMarkdown": "Thank you so much ! It's just as you say, in my case, I learned a lot from each competition, thanks to the invaluable discussions and the code shared by the kagglers. I think this is the most valuable thing that competitions leave us...beyond the medals that always make us happy :)", "votes": null } ], "comments": [ { "id": 3076563, "author_name": "yekenot", "author_url": "", "post_date": "12/20/2024 04:30:48", "content": "Congratulations. I was also inspired by this winning DNN but used another one. IterativeImputer produced good results in my pipeline as well; I preferred this technique among other imputation ways.
Felicitaciones Laura! Increíble shakeup, gran trabajo. Buen autoregalo de navidad la medalla de oro ja. Felices fiestas!
", "votes": null, "replies": [ { "id": 3077263, "author_name": "lauraromar", "author_url": "", "post_date": "12/20/2024 18:07:00", "content": "Muchas gracias Maxi ! Muy contenta con mi autoregalito Navideño, jaja. Felices fiestas !
", "votes": null, "replies": [] } ] }, { "id": 3078698, "author_name": "saidkoussi", "author_url": "", "post_date": "12/22/2024 17:23:58", "content": "@lauraromar firstly congratulation , isee that you participate in ICR competition you are not get good position in it but it seems that you learn a lot from it, what i learn from your post is that when you work hard in a competition you never lose , if you get gold or silver medal that good and if you get nothing you'll be prepared fr next competitions what is important is to keep going
", "votes": null, "replies": [ { "id": 3078882, "author_name": "lauraromar", "author_url": "", "post_date": "12/23/2024 01:21:47", "content": "Thank you so much ! It's just as you say, in my case, I learned a lot from each competition, thanks to the invaluable discussions and the code shared by the kagglers. I think this is the most valuable thing that competitions leave us…beyond the medals that always make us happy :)
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "3076558": "Hello everyone! I am very happy to have won a gold medal and I would like to share my solution with you:\n- This is a single classification model (yes! not a regression model)\n- This solution was based on some sophisticated DNN that was the winning solution of the ICR competition. I loved it when I saw it and I was waiting for a competition where I could apply it.\n- Use IterativeImputer class as a strategy for imputing the missing values, which models each feature with missing values based on other features and uses that estimate for imputation. I decided to use both the train and test data to perform the fit and trained the model online so that I could take the data from the private test.\n- Take the enmo from the actigraphy data differentiating weekdays from weekends and grouping them by day and time slots.\n- Train with ALL the data\n- Use weights to balance the classes in the loss function\n- Take the average of the predictions of the same model with 10 different seeds\n\nMany thanks to the organizers of this competition and to everyone who participated by sharing their ideas and code !\n\nHere I share the code of my solution.", "3076563": "Congratulations. I was also inspired by this winning DNN but used another one. `IterativeImputer` produced good results in my pipeline as well; I preferred this technique among other imputation ways.", "3076843": "Felicitaciones Laura! Increíble shakeup, gran trabajo. Buen autoregalo de navidad la medalla de oro ja. Felices fiestas!", "3077263": "Muchas gracias Maxi ! Muy contenta con mi autoregalito Navideño, jaja. Felices fiestas !", "3078698": "lauraromar firstly congratulation , isee that you participate in ICR competition you are not get good position in it but it seems that you learn a lot from it, what i learn from your post is that when you work hard in a competition you never lose , if you get gold or silver medal that good and if you get nothing you'll be prepared fr next competitions what is important is to keep going", "3078882": "Thank you so much ! It's just as you say, in my case, I learned a lot from each competition, thanks to the invaluable discussions and the code shared by the kagglers. I think this is the most valuable thing that competitions leave us...beyond the medals that always make us happy :)" }, "source": "meta" }