{
  "id": 582917,
  "title": "How can we check for outliers in time series data?",
  "url": "/competitions/cmi-detect-behavior-with-sensor-data/discussion/582917",
  "author_name": "",
  "post_date": "2025-06-03T16:57:14.538288500Z",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hi everyone, i am new to data involving time series. How do we analyse for outliers here? I came across the idea plotting the whole acc_x column, but doesn't that include all the timesteps from all the sequences, and values vary with a pattern over timesteps, so how to analyse the distribution effectively?</p>",
  "messages": [
    {
      "id": "3216505",
      "postDate": "06/03/2025 16:57:14",
      "content": "<p>Hi everyone, i am new to data involving time series. How do we analyse for outliers here? I came across the idea plotting the whole acc_x column, but doesn't that include all the timesteps from all the sequences, and values vary with a pattern over timesteps, so how to analyse the distribution effectively?</p>",
      "rawMarkdown": "Hi everyone, i am new to data involving time series. How do we analyse for outliers here? I came across the idea plotting the whole acc_x column, but doesn't that include all the timesteps from all the sequences, and values vary with a pattern over timesteps, so how to analyse the distribution effectively?",
      "votes": null
    },
    {
      "id": "3216598",
      "postDate": "06/03/2025 19:43:06",
      "content": "<p><a href=\"https://www.kaggle.com/abhirupchoudhury073\" target=\"_blank\">@abhirupchoudhury073</a>  This isn't limited to time-series data, but if we standardize the data using (X-mean)/std, wouldn't it be sufficient to then just look at its minimum and maximum values?</p>",
      "rawMarkdown": "abhirupchoudhury073  This isn't limited to time-series data, but if we standardize the data using (X-mean)/std, wouldn't it be sufficient to then just look at its minimum and maximum values?",
      "votes": null
    },
    {
      "id": "3217417",
      "postDate": "06/05/2025 02:07:28",
      "content": "<p><a href=\"https://www.kaggle.com/abhirupchoudhury073\" target=\"_blank\">@abhirupchoudhury073</a> , use fourier transform and see the data in frequency domain may be a great idea</p>",
      "rawMarkdown": "abhirupchoudhury073 , use fourier transform and see the data in frequency domain may be a great idea",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3216598,
      "author_name": "chumajin",
      "author_url": "",
      "post_date": "06/03/2025 19:43:06",
      "content": "<p><a href=\"https://www.kaggle.com/abhirupchoudhury073\" target=\"_blank\">@abhirupchoudhury073</a>  This isn't limited to time-series data, but if we standardize the data using (X-mean)/std, wouldn't it be sufficient to then just look at its minimum and maximum values?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 3217417,
      "author_name": "christina0626",
      "author_url": "",
      "post_date": "06/05/2025 02:07:28",
      "content": "<p><a href=\"https://www.kaggle.com/abhirupchoudhury073\" target=\"_blank\">@abhirupchoudhury073</a> , use fourier transform and see the data in frequency domain may be a great idea</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3216505": "Hi everyone, i am new to data involving time series. How do we analyse for outliers here? I came across the idea plotting the whole acc_x column, but doesn't that include all the timesteps from all the sequences, and values vary with a pattern over timesteps, so how to analyse the distribution effectively?",
    "3216598": "abhirupchoudhury073  This isn't limited to time-series data, but if we standardize the data using (X-mean)/std, wouldn't it be sufficient to then just look at its minimum and maximum values?",
    "3217417": "abhirupchoudhury073 , use fourier transform and see the data in frequency domain may be a great idea"
  },
  "source": "meta"
}