{
  "id": 388298,
  "title": "[Anomaly Thread] Those sessions are wierd..",
  "url": "/competitions/predict-student-performance-from-game-play/discussion/388298",
  "author_name": "",
  "post_date": "2023-02-16T20:23:03.407906900Z",
  "votes": 2,
  "comment_count": 3,
  "views": 0,
  "content": "<p>see <a href=\"https://www.kaggle.com/code/janmpia/student-perf-eda-feature-engineering\" target=\"_blank\">here </a> for the df_sessions dataset -&gt;</p>\n<pre><code>df_sessions[] = df_sessions.num_events / df_sessions.last_elapsed_time\ndf_sessions[df_sessions.eps == df_sessions.eps.()][[,,,,]]\n</code></pre>\n<p>|num_events |accuracy|last_elapsed_time|eps|</p>\n<table>\n<thead>\n<tr>\n<th>session_id</th>\n<th>num_events</th>\n<th>accuracy</th>\n<th>last_elapsed_time</th>\n<th>eps</th>\n</tr>\n</thead>\n<tbody>\n<tr>\n<td>21020618143279870</td>\n<td>718</td>\n<td>0.944444</td>\n<td>2</td>\n<td>359.0</td>\n</tr>\n</tbody>\n</table>\n<p>this session lasted <strong>2 seconds and the user has done 718 event during that time</strong>, giving a event/second of 359.0, which doesn't make much sense, the worst part is that he still got <strong>0.944444 question accuracy</strong>.</p>\n<p>Might be worth <strong>taking that session_id of your model to get better predictions.</strong><br>\nThx for sharing other anomalies in the Dataset !</p>",
  "messages": [
    {
      "id": "2147721",
      "postDate": "02/16/2023 20:23:03",
      "content": "<p>see <a href=\"https://www.kaggle.com/code/janmpia/student-perf-eda-feature-engineering\" target=\"_blank\">here </a> for the df_sessions dataset -&gt;</p>\n<pre><code>df_sessions[] = df_sessions.num_events / df_sessions.last_elapsed_time\ndf_sessions[df_sessions.eps == df_sessions.eps.()][[,,,,]]\n</code></pre>\n<p>|num_events |accuracy|last_elapsed_time|eps|</p>\n<table>\n<thead>\n<tr>\n<th>session_id</th>\n<th>num_events</th>\n<th>accuracy</th>\n<th>last_elapsed_time</th>\n<th>eps</th>\n</tr>\n</thead>\n<tbody>\n<tr>\n<td>21020618143279870</td>\n<td>718</td>\n<td>0.944444</td>\n<td>2</td>\n<td>359.0</td>\n</tr>\n</tbody>\n</table>\n<p>this session lasted <strong>2 seconds and the user has done 718 event during that time</strong>, giving a event/second of 359.0, which doesn't make much sense, the worst part is that he still got <strong>0.944444 question accuracy</strong>.</p>\n<p>Might be worth <strong>taking that session_id of your model to get better predictions.</strong><br>\nThx for sharing other anomalies in the Dataset !</p>",
      "rawMarkdown": "see [here ](https://www.kaggle.com/code/janmpia/student-perf-eda-feature-engineering) for the df_sessions dataset ->\n```python\n\ndf_sessions['eps'] = df_sessions.num_events / df_sessions.last_elapsed_time\ndf_sessions[df_sessions.eps == df_sessions.eps.max()][['session_id','num_events','accuracy','last_elapsed_time','eps']]\n```\n\n|num_events |accuracy|last_elapsed_time|eps|\n\n|session_id|num_events |accuracy|last_elapsed_time|eps|\n| --- | --- |\n| 21020618143279870| 718|0.944444|2|359.0|\n\nthis session lasted **2 seconds and the user has done 718 event during that time**, giving a event/second of 359.0, which doesn't make much sense, the worst part is that he still got **0.944444 question accuracy**.\n\nMight be worth **taking that session_id of your model to get better predictions.**\nThx for sharing other anomalies in the Dataset !",
      "votes": null
    },
    {
      "id": "2147982",
      "postDate": "02/17/2023 04:26:21",
      "content": "<p>Thanks for pointing this out! Do you know how many users have such abnormally high events/sec (I'd say anything above 10 is irregular)?</p>",
      "rawMarkdown": "Thanks for pointing this out! Do you know how many users have such abnormally high events/sec (I'd say anything above 10 is irregular)?",
      "votes": null
    },
    {
      "id": "2148247",
      "postDate": "02/17/2023 09:17:02",
      "content": "<p>thx ! just checked and thats the only session with above 10 eps !</p>",
      "rawMarkdown": "thx ! just checked and thats the only session with above 10 eps !",
      "votes": null
    },
    {
      "id": "2148251",
      "postDate": "02/17/2023 09:24:29",
      "content": "<p>Glad to hear - thanks!</p>",
      "rawMarkdown": "Glad to hear - thanks!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2147982,
      "author_name": "hoangnguyen719",
      "author_url": "",
      "post_date": "02/17/2023 04:26:21",
      "content": "<p>Thanks for pointing this out! Do you know how many users have such abnormally high events/sec (I'd say anything above 10 is irregular)?</p>",
      "votes": null,
      "replies": [
        {
          "id": 2148247,
          "author_name": "janmpia",
          "author_url": "",
          "post_date": "02/17/2023 09:17:02",
          "content": "<p>thx ! just checked and thats the only session with above 10 eps !</p>",
          "votes": null,
          "replies": [
            {
              "id": 2148251,
              "author_name": "hoangnguyen719",
              "author_url": "",
              "post_date": "02/17/2023 09:24:29",
              "content": "<p>Glad to hear - thanks!</p>",
              "votes": null,
              "replies": []
            }
          ]
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2147721": "see [here ](https://www.kaggle.com/code/janmpia/student-perf-eda-feature-engineering) for the df_sessions dataset ->\n```python\n\ndf_sessions['eps'] = df_sessions.num_events / df_sessions.last_elapsed_time\ndf_sessions[df_sessions.eps == df_sessions.eps.max()][['session_id','num_events','accuracy','last_elapsed_time','eps']]\n```\n\n|num_events |accuracy|last_elapsed_time|eps|\n\n|session_id|num_events |accuracy|last_elapsed_time|eps|\n| --- | --- |\n| 21020618143279870| 718|0.944444|2|359.0|\n\nthis session lasted **2 seconds and the user has done 718 event during that time**, giving a event/second of 359.0, which doesn't make much sense, the worst part is that he still got **0.944444 question accuracy**.\n\nMight be worth **taking that session_id of your model to get better predictions.**\nThx for sharing other anomalies in the Dataset !",
    "2147982": "Thanks for pointing this out! Do you know how many users have such abnormally high events/sec (I'd say anything above 10 is irregular)?",
    "2148247": "thx ! just checked and thats the only session with above 10 eps !",
    "2148251": "Glad to hear - thanks!"
  },
  "source": "meta"
}