\n
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F12062233%2F2924ed040bcb38a63b1b57eea1bcebd9%2Fsolution.jpg?generation=1741267861031260&alt=media\")
First, congratulations to the competition winners, and sincere thanks to our team members @gentlezdh, @yoshifumimiya, @lizhecheng, @nandodmelo, @kekshibata , as well as the competition organizers.
\nWe present a short summary of our solution (24th place, private LB 0.694). More detailed version is here.
We trained regressors and classifiers separately and combined their outputs using ensembling & post-processing. And the “Trust CV” strategy led us to a final LB score of 0.694 to win silver.
\n
Selected the best CV and LB for final submissions. “Trust CV” approach ultimately secured us the silver medal.
", "comments": [ { "id": 3142587, "author_name": "Donghui Zhang", "author_url": "", "post_date": "2025-03-06T14:00:51.937000", "content": "Enjoy this competition with our teammates!
", "votes": 4, "replies": [] }, { "id": 3142573, "author_name": "Octopus210", "author_url": "", "post_date": "2025-03-06T13:41:51.090000", "content": "I also appreciate you all！
\nYou are a great engineer😆
First, congratulations to the competition winners, and sincere thanks to our team members @gentlezdh, @yoshifumimiya, @lizhecheng, @nandodmelo, @kekshibata , as well as the competition organizers.
\nWe present a short summary of our solution (24th place, private LB 0.694). More detailed version is here.
We trained regressors and classifiers separately and combined their outputs using ensembling & post-processing. And the “Trust CV” strategy led us to a final LB score of 0.694 to win silver.
\n
Selected the best CV and LB for final submissions. “Trust CV” approach ultimately secured us the silver medal.
", "rawMarkdown": "First, congratulations to the competition winners, and sincere thanks to our team members @gentlezdh, @yoshifumimiya, @lizhecheng, @nandodmelo, @kekshibata , as well as the competition organizers.\nWe present a short summary of our solution (**24th place, private LB 0.694**). More detailed version is [here](https://www.kaggle.com/competitions/equity-post-HCT-survival-predictions/discussion/566809).\n\n# Overview\nWe trained regressors and classifiers separately and combined their outputs using ensembling & post-processing. And the “Trust CV” strategy led us to a final LB score of 0.694 to win silver.\n\n\n# Model Training\n## Regressors:\n- GBDTs (CAT, LGB, XGB) and AutoGluon on 8 targets with basic feature engineering.\n- For specific 4 targets, applied extensive feature engineering with CatBoost and AutoGluon\n- PRL NNs, including customized losses\n\n## Classifiers:\n- GBDT (CAT, LGB, XGB) classifiers\n- Stacked logistic regression and combined Autogluon classifier.\n\n# Ensemble & Post-Processing\n- Combined regressors’ outputs using a **multi-task Elastic Net** on 8 targets. (excluding some custom PRL NNs as inputs).\n- Post-processing: Adjusted predictions based on classifier probabilities with race-specific parameters optimized by Optuna.\n- Blended in custom PRL NNs for high efs probability data and weaker races.\n\n# Submission Strategy\n- Best CV: CV 0.6908 / public LB 0.690 / private LB 0.694\n- Best LB: CV 0.6853 / public LB 0.694 / private LB 0.692\n\nSelected the best CV and LB for final submissions. “Trust CV” approach ultimately secured us the silver medal.\n", "votes": 9 }, { "id": 3142587, "postDate": "2025-03-06T14:00:51.937Z", "content": "Enjoy this competition with our teammates!
", "rawMarkdown": "Enjoy this competition with our teammates!", "votes": 4 }, { "id": 3142573, "postDate": "2025-03-06T13:41:51.090Z", "content": "I also appreciate you all！
\nYou are a great engineer😆