{
  "id": 599699,
  "title": "Missing Sensors in the Test Data",
  "url": "/competitions/cmi-detect-behavior-with-sensor-data/discussion/599699",
  "author_name": "",
  "post_date": "2025-08-18T08:48:49.856282200Z",
  "votes": 2,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I have a question, you have mentioned that half of the data in the competition have only the IMU sensor data, the test.csv file contains the whole sensors, I know that the leaderboard score uses only 42% of the data, but I want to know what if the perdict method does not handle the situation of missing data in the test set, does the submission will be treated as failed or how will you deal with this?</p>",
  "messages": [
    {
      "id": "3271180",
      "postDate": "08/18/2025 08:48:49",
      "content": "<p>I have a question, you have mentioned that half of the data in the competition have only the IMU sensor data, the test.csv file contains the whole sensors, I know that the leaderboard score uses only 42% of the data, but I want to know what if the perdict method does not handle the situation of missing data in the test set, does the submission will be treated as failed or how will you deal with this?</p>",
      "rawMarkdown": "I have a question, you have mentioned that half of the data in the competition have only the IMU sensor data, the test.csv file contains the whole sensors, I know that the leaderboard score uses only 42% of the data, but I want to know what if the perdict method does not handle the situation of missing data in the test set, does the submission will be treated as failed or how will you deal with this?",
      "votes": null
    },
    {
      "id": "3273910",
      "postDate": "08/23/2025 16:06:23",
      "content": "<p>I believe there are non nans, rather there are 0 values for the sensor that have been \"removed\".<br>\nHowever if your predict function raises an error during submission you won't get score.<br>\nSo I would suggest using a safeguard fillna call inside your preprocessing.</p>",
      "rawMarkdown": "I believe there are non nans, rather there are 0 values for the sensor that have been \"removed\".\nHowever if your predict function raises an error during submission you won't get score.\nSo I would suggest using a safeguard fillna call inside your preprocessing.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3273910,
      "author_name": "mauroabidalcarrer",
      "author_url": "",
      "post_date": "08/23/2025 16:06:23",
      "content": "<p>I believe there are non nans, rather there are 0 values for the sensor that have been \"removed\".<br>\nHowever if your predict function raises an error during submission you won't get score.<br>\nSo I would suggest using a safeguard fillna call inside your preprocessing.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3271180": "I have a question, you have mentioned that half of the data in the competition have only the IMU sensor data, the test.csv file contains the whole sensors, I know that the leaderboard score uses only 42% of the data, but I want to know what if the perdict method does not handle the situation of missing data in the test set, does the submission will be treated as failed or how will you deal with this?",
    "3273910": "I believe there are non nans, rather there are 0 values for the sensor that have been \"removed\".\nHowever if your predict function raises an error during submission you won't get score.\nSo I would suggest using a safeguard fillna call inside your preprocessing."
  },
  "source": "meta"
}