![\"Overview\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2848721%2F279f932e866a8efdd2c539776f6436e6%2F2023-05-19%209.56.02.png?generation=1684460061564118&alt=media\")
Before anything else, I want to express my deep gratitude to my team members @vitalykudelya @yukisawamura @yukisawamura @salaryman who contributed their time, efforts, and expertise in this competition. Next, we'd like to extend our thanks to the Kaggle community and the hosts. Thank you for providing a challenging dataset and a great learning opportunity.
\n
Our solution is based on user grouping, trend calculation for each group, user group prediction (classification), and trend assignment for each group.
\nYou can check our solution code here. It takes super long time to be scored beacuse my pipeline is not good😓
During our exploratory data analysis (EDA), we noticed significant differences in the target trends among users based on the presence or absence of Medication information (across all 'visit_month' data, regardless of being 'On' or 'Off'). It is basis of our pipeline.
\nBased on the groups' information, we created three types of trends: those without Medication information, those with Medication information, and overall trends. These trends were adopted from the 'Only_Trend' and 'Protein Shift' notebooks. Particularly, the protein 'P05060' greatly contributed to improving the Public LB score.
\nWe implemented binary classification using LightGBM to predict whether a user belongs to a group with Medication information or not. The prediction was made based on 'visit_month' related information, and Protein and Peptide data.
\nFor our final submission, to increase the robustness of the model, we utilized 10-fold cross-validation (CV) and random seed averaging.
Based on the results of the binary classification, we mapped the trends of each group. For our final submission, if the prediction from binary classification was above 0.75, we assumed it belonged to the group with Medication information. If it was below 0.15, we assumed it belonged to the group without Medication information. For all other cases, we used the overall trend.
\nAs the dataset for this competition was small, both CV and LB scores were unstable. We struggled until the end to decide which submission to choose. If you have any questions or feedback, please feel free to comment. We look forward to learning with all of you in the community.
", "comments": [ { "id": 2265160, "author_name": "IonDrive", "author_url": "", "post_date": "2023-05-19T03:12:45.063000", "content": "Congratulations Vitaly and Stallone on becoming competition masters! And as well to Maruichi and Salaryman on becoming competition experts! Thanks for sharing your solution, as well as the public solutions you provided during the contest.
", "votes": 3, "replies": [] }, { "id": 2265119, "author_name": "AC", "author_url": "", "post_date": "2023-05-19T02:03:06.383000", "content": "Hey! Thank you for your notebook! And Congratulations! :)
\nDo you know why this particular protein 'P05060' is giving better results. I submitted a notebook after reading this post of yours right now and my public lb score got a boost from 55.5 to 54.9 and even my private lb improved.
P05060 was chosen as the best protein improving the score on the train dataset (for NPX groups top5 quantile and low5 quantile) without any medical knowledge
\nProtein P05060 imporoved cross-validation score, public score and private score over Trend.
\nI'm not sure if this is pure luck or if we have a real signal in the P05060 protein )
Secretogranin-1 is a neuroendocrine secretory granule protein, which may be the precursor for other biologically active peptides
", "votes": 2, "replies": [] } ] } ], "comment_count": 4, "messages": [ { "id": 2265111, "postDate": "2023-05-19T01:48:48.830Z", "content": "Before anything else, I want to express my deep gratitude to my team members @vitalykudelya @yukisawamura @yukisawamura @salaryman who contributed their time, efforts, and expertise in this competition. Next, we'd like to extend our thanks to the Kaggle community and the hosts. Thank you for providing a challenging dataset and a great learning opportunity.
\n
Our solution is based on user grouping, trend calculation for each group, user group prediction (classification), and trend assignment for each group.
\nYou can check our solution code here. It takes super long time to be scored beacuse my pipeline is not good😓
During our exploratory data analysis (EDA), we noticed significant differences in the target trends among users based on the presence or absence of Medication information (across all 'visit_month' data, regardless of being 'On' or 'Off'). It is basis of our pipeline.
\nBased on the groups' information, we created three types of trends: those without Medication information, those with Medication information, and overall trends. These trends were adopted from the 'Only_Trend' and 'Protein Shift' notebooks. Particularly, the protein 'P05060' greatly contributed to improving the Public LB score.
\nWe implemented binary classification using LightGBM to predict whether a user belongs to a group with Medication information or not. The prediction was made based on 'visit_month' related information, and Protein and Peptide data.
\nFor our final submission, to increase the robustness of the model, we utilized 10-fold cross-validation (CV) and random seed averaging.
Based on the results of the binary classification, we mapped the trends of each group. For our final submission, if the prediction from binary classification was above 0.75, we assumed it belonged to the group with Medication information. If it was below 0.15, we assumed it belonged to the group without Medication information. For all other cases, we used the overall trend.
\nAs the dataset for this competition was small, both CV and LB scores were unstable. We struggled until the end to decide which submission to choose. If you have any questions or feedback, please feel free to comment. We look forward to learning with all of you in the community.
", "rawMarkdown": "Before anything else, I want to express my deep gratitude to my team members @vitalykudelya @yukisawamura @yukisawamura @salaryman who contributed their time, efforts, and expertise in this competition. Next, we'd like to extend our thanks to the Kaggle community and the hosts. Thank you for providing a challenging dataset and a great learning opportunity.\n\n# **Overview of our solution**\n\n\nOur solution is based on user grouping, trend calculation for each group, user group prediction (classification), and trend assignment for each group.\nYou can check our solution code [here](https://www.kaggle.com/maruichi01/11th-place-solution). It takes super long time to be scored beacuse my pipeline is not good😓\n\n## **1. Grouping**\nDuring our exploratory data analysis (EDA), we noticed significant differences in the target trends among users based on the presence or absence of Medication information (across all 'visit_month' data, regardless of being 'On' or 'Off'). It is basis of our pipeline.\n\n## **2. Trend Calculation**\nBased on the groups' information, we created three types of trends: those without Medication information, those with Medication information, and overall trends. These trends were adopted from the '[Only_Trend](https://www.kaggle.com/code/vitalykudelya/only-trends)' and '[Protein Shift](https://www.kaggle.com/code/vitalykudelya/p05060-protein-npx-groups-trend-silver-medal)' notebooks. Particularly, the protein 'P05060' greatly contributed to improving the Public LB score.\n\n## **3. Feature Engineering + Group Classification**\nWe implemented binary classification using LightGBM to predict whether a user belongs to a group with Medication information or not. The prediction was made based on 'visit_month' related information, and Protein and Peptide data.\nFor our final submission, to increase the robustness of the model, we utilized 10-fold cross-validation (CV) and random seed averaging.\n\n## **4. Using Binary Classification Results for Trend Mapping**\nBased on the results of the binary classification, we mapped the trends of each group. For our final submission, if the prediction from binary classification was above 0.75, we assumed it belonged to the group with Medication information. If it was below 0.15, we assumed it belonged to the group without Medication information. For all other cases, we used the overall trend.\n\nAs the dataset for this competition was small, both CV and LB scores were unstable. We struggled until the end to decide which submission to choose. If you have any questions or feedback, please feel free to comment. We look forward to learning with all of you in the community.", "votes": 19 }, { "id": 2265160, "postDate": "2023-05-19T03:12:45.063Z", "content": "Congratulations Vitaly and Stallone on becoming competition masters! And as well to Maruichi and Salaryman on becoming competition experts! Thanks for sharing your solution, as well as the public solutions you provided during the contest.
", "rawMarkdown": "Congratulations Vitaly and Stallone on becoming competition masters! And as well to Maruichi and Salaryman on becoming competition experts! Thanks for sharing your solution, as well as the public solutions you provided during the contest.", "votes": 3 }, { "id": 2265119, "postDate": "2023-05-19T02:03:06.383Z", "content": "Hey! Thank you for your notebook! And Congratulations! :)
\nDo you know why this particular protein 'P05060' is giving better results. I submitted a notebook after reading this post of yours right now and my public lb score got a boost from 55.5 to 54.9 and even my private lb improved.
P05060 was chosen as the best protein improving the score on the train dataset (for NPX groups top5 quantile and low5 quantile) without any medical knowledge
\nProtein P05060 imporoved cross-validation score, public score and private score over Trend.
\nI'm not sure if this is pure luck or if we have a real signal in the P05060 protein )
Secretogranin-1 is a neuroendocrine secretory granule protein, which may be the precursor for other biologically active peptides
", "rawMarkdown": "Secretogranin-1 is a neuroendocrine secretory granule protein, which may be the precursor for other biologically active peptides", "votes": 2 } ] } ], "complete": true, "source": "live", "fetched_at": "2026-10-19T14:11:25Z", "competition_id": 37794, "competition_slug": "amp-parkinsons-disease-progression-prediction", "team_id": 10219384, "solution_url": "/competitions/amp-parkinsons-disease-progression-prediction/writeups/win-kinakomochi-vs-sophia-12th-place-solution" }