{ "id": 411395, "title": "8th place solution: One trick to win the gold", "content": "

Thanks my teammates @hookman @zui0711 @max2020 @librauee , I have learnt a lot from them. A crucial trick found by them is there are two groups in the data: The true patients and the control group. We can find this by the minimum visit_month diff of each id:

\n
for df in [clinical, sup_clinical]:\n    df['visit_month_diff'] = df.groupby(['patient_id'])['visit_month'].diff()\n    df['visit_month_diff_min'] = df.groupby('patient_id')['visit_month_diff'].transform('min')\n
\n

There are 3, 5, 6, 12, 36 visit_month diff in clinical and sup_clinical, we choose only 3, 6, 12, 36 parts as training data, and transform 3, 36 to 6. We find 3, 6 and 36 has obviously higher updrs values and 12 has lower updrs values(5 part dropped as abnormal).
\n\"updrs
\nWe can draw a conclusion visit_month diff==6 are true patients and visit_month diff==12 are control group.
\nWe use piecewise function to optimize two groups smape separately:

\n
def calculate_predictions(pred_month, trend):\n    if target == 'updrs_4': \n        pred_month = pred_month.clip(60, None) \n    pred_month2 = (pred_month-60).clip(0, None)\n    return np.round(trend[0] + pred_month * trend[1] + pred_month2 * trend[2])\n
\n

\"0=control
\nIn the first loop inference, we cannot get visit_month diff, so we train additional coefficients based on true patients + control group. In every loop, we record history data so that we could get the the minimum visit_month diff.
\nFinding these two groups can help you reach 54.2~54.8 on public board and 60.1~60.7 on private board.

", "comments": [], "comment_count": 0, "messages": [ { "id": 2265120, "postDate": "2023-05-19T02:03:15.737Z", "content": "

Thanks my teammates @hookman @zui0711 @max2020 @librauee , I have learnt a lot from them. A crucial trick found by them is there are two groups in the data: The true patients and the control group. We can find this by the minimum visit_month diff of each id:

\n
for df in [clinical, sup_clinical]:\n    df['visit_month_diff'] = df.groupby(['patient_id'])['visit_month'].diff()\n    df['visit_month_diff_min'] = df.groupby('patient_id')['visit_month_diff'].transform('min')\n
\n

There are 3, 5, 6, 12, 36 visit_month diff in clinical and sup_clinical, we choose only 3, 6, 12, 36 parts as training data, and transform 3, 36 to 6. We find 3, 6 and 36 has obviously higher updrs values and 12 has lower updrs values(5 part dropped as abnormal).
\n\"updrs
\nWe can draw a conclusion visit_month diff==6 are true patients and visit_month diff==12 are control group.
\nWe use piecewise function to optimize two groups smape separately:

\n
def calculate_predictions(pred_month, trend):\n    if target == 'updrs_4': \n        pred_month = pred_month.clip(60, None) \n    pred_month2 = (pred_month-60).clip(0, None)\n    return np.round(trend[0] + pred_month * trend[1] + pred_month2 * trend[2])\n
\n

\"0=control
\nIn the first loop inference, we cannot get visit_month diff, so we train additional coefficients based on true patients + control group. In every loop, we record history data so that we could get the the minimum visit_month diff.
\nFinding these two groups can help you reach 54.2~54.8 on public board and 60.1~60.7 on private board.

", "rawMarkdown": "Thanks my teammates @hookman @zui0711 @max2020 @librauee , I have learnt a lot from them. A crucial trick found by them is there are two groups in the data: **The true patients and the control group**. We can find this by **the minimum visit_month diff** of each id:\n```\nfor df in [clinical, sup_clinical]:\n df['visit_month_diff'] = df.groupby(['patient_id'])['visit_month'].diff()\n df['visit_month_diff_min'] = df.groupby('patient_id')['visit_month_diff'].transform('min')\n```\nThere are 3, 5, 6, 12, 36 visit_month diff in clinical and sup_clinical, we choose only 3, 6, 12, 36 parts as training data, and transform 3, 36 to 6. We find 3, 6 and 36 has obviously higher updrs values and 12 has lower updrs values(5 part dropped as abnormal).\n![updrs in different visit_month diff and months](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F3110858%2F99690a48cfd9b72ad8bc4fc746815856%2FWX20230519-0926512x.png?generation=1684460243462016&alt=media)\nWe can draw a conclusion **visit_month diff==6** are true patients and **visit_month diff==12** are control group.\nWe use piecewise function to optimize two groups smape separately:\n```\ndef calculate_predictions(pred_month, trend):\n if target == 'updrs_4': \n pred_month = pred_month.clip(60, None) \n pred_month2 = (pred_month-60).clip(0, None)\n return np.round(trend[0] + pred_month * trend[1] + pred_month2 * trend[2])\n```\n![0=control group, 1=true patients](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F3110858%2F5705a4149bab70f274bd16c521c5c967%2FWX20230519-1027572x.png?generation=1684463307465921&alt=media)\nIn the first loop inference, we cannot get visit_month diff, so we train additional coefficients based on true patients + control group. In every loop, we record history data so that we could get the **the minimum visit_month diff**.\nFinding these two groups can help you reach 54.2~54.8 on public board and 60.1~60.7 on private board.", "votes": 28 } ], "complete": true, "source": "live", "fetched_at": "2026-10-19T14:12:03Z", "competition_id": 37794, "competition_slug": "amp-parkinsons-disease-progression-prediction", "team_id": 9913452, "solution_url": "/competitions/amp-parkinsons-disease-progression-prediction/writeups/all-luck-no-skill-8th-place-solution-one-trick-to-" }