{
  "id": 318426,
  "title": "Time feature ? ",
  "url": "/competitions/geolifeclef-2022-lifeclef-2022-fgvc9/discussion/318426",
  "author_name": "",
  "post_date": "2022-04-12T08:39:40.132125200Z",
  "votes": 3,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I was wondering if the time of the observation would matter for such a problem. I can see at least two ways time would be a significant driver:</p>\n<pre><code>- 1) time passing by: if the demographic is changing fast, a specie going extinct for exemple. Predicing this specie for time past extinction wouldn't make much sense.\n- 2) time of the year (season): we could avoid predicting some migrating species in the winter for exemple... probably more impacting on the animalia kingdom. But I wouldn't be surprised some small plants are way less visible in winter too. \n</code></pre>\n<p>Time is unfortunately not available from data. I tried some approaches to get a proxy for it without sucess. But I got more questions than answer from it. Typically: shouldn't the environement change depending on season ? </p>\n<p>What are your toughts on time as a feature ?</p>",
  "messages": [
    {
      "id": "1752875",
      "postDate": "04/12/2022 08:39:40",
      "content": "<p>I was wondering if the time of the observation would matter for such a problem. I can see at least two ways time would be a significant driver:</p>\n<pre><code>- 1) time passing by: if the demographic is changing fast, a specie going extinct for exemple. Predicing this specie for time past extinction wouldn't make much sense.\n- 2) time of the year (season): we could avoid predicting some migrating species in the winter for exemple... probably more impacting on the animalia kingdom. But I wouldn't be surprised some small plants are way less visible in winter too. \n</code></pre>\n<p>Time is unfortunately not available from data. I tried some approaches to get a proxy for it without sucess. But I got more questions than answer from it. Typically: shouldn't the environement change depending on season ? </p>\n<p>What are your toughts on time as a feature ?</p>",
      "rawMarkdown": "I was wondering if the time of the observation would matter for such a problem. I can see at least two ways time would be a significant driver:\n\n    - 1) time passing by: if the demographic is changing fast, a specie going extinct for exemple. Predicing this specie for time past extinction wouldn't make much sense.\n    - 2) time of the year (season): we could avoid predicting some migrating species in the winter for exemple... probably more impacting on the animalia kingdom. But I wouldn't be surprised some small plants are way less visible in winter too. \n\nTime is unfortunately not available from data. I tried some approaches to get a proxy for it without sucess. But I got more questions than answer from it. Typically: shouldn't the environement change depending on season ? \n\nWhat are your toughts on time as a feature ?",
      "votes": null
    },
    {
      "id": "1795103",
      "postDate": "05/19/2022 12:09:04",
      "content": "<p>Hi Lucas,</p>\n<p>Sorry for the late reply, I forgot to answer before…</p>\n<p>That's some very good points, they are all true.<br>\nHowever, their relevance depends on the context and what we want to model.</p>\n<p>In the case of GeoLifeCLEF 2022, we are interested in knowing where a species can be observed at some point of the year (without specifying it), e.g., all migrating species passing by, annual plants observable only at a time of the year, etc. That's why we don't give the time of the year.<br>\nFor the \"time passing by\", in fact all the observations were collected in a single year (2019). Geographical shift of species distributions (due to extinctions or others) occur on a larger time scale and are thus not detectable here.</p>\n<p>In any case, I don't think you will be able to recover part of this time information using the provided data.</p>",
      "rawMarkdown": "Hi Lucas,\n\nSorry for the late reply, I forgot to answer before...\n\nThat's some very good points, they are all true.\nHowever, their relevance depends on the context and what we want to model.\n\nIn the case of GeoLifeCLEF 2022, we are interested in knowing where a species can be observed at some point of the year (without specifying it), e.g., all migrating species passing by, annual plants observable only at a time of the year, etc. That's why we don't give the time of the year.\nFor the \"time passing by\", in fact all the observations were collected in a single year (2019). Geographical shift of species distributions (due to extinctions or others) occur on a larger time scale and are thus not detectable here.\n\nIn any case, I don't think you will be able to recover part of this time information using the provided data.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1795103,
      "author_name": "tlorieul",
      "author_url": "",
      "post_date": "05/19/2022 12:09:04",
      "content": "<p>Hi Lucas,</p>\n<p>Sorry for the late reply, I forgot to answer before…</p>\n<p>That's some very good points, they are all true.<br>\nHowever, their relevance depends on the context and what we want to model.</p>\n<p>In the case of GeoLifeCLEF 2022, we are interested in knowing where a species can be observed at some point of the year (without specifying it), e.g., all migrating species passing by, annual plants observable only at a time of the year, etc. That's why we don't give the time of the year.<br>\nFor the \"time passing by\", in fact all the observations were collected in a single year (2019). Geographical shift of species distributions (due to extinctions or others) occur on a larger time scale and are thus not detectable here.</p>\n<p>In any case, I don't think you will be able to recover part of this time information using the provided data.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1752875": "I was wondering if the time of the observation would matter for such a problem. I can see at least two ways time would be a significant driver:\n\n    - 1) time passing by: if the demographic is changing fast, a specie going extinct for exemple. Predicing this specie for time past extinction wouldn't make much sense.\n    - 2) time of the year (season): we could avoid predicting some migrating species in the winter for exemple... probably more impacting on the animalia kingdom. But I wouldn't be surprised some small plants are way less visible in winter too. \n\nTime is unfortunately not available from data. I tried some approaches to get a proxy for it without sucess. But I got more questions than answer from it. Typically: shouldn't the environement change depending on season ? \n\nWhat are your toughts on time as a feature ?",
    "1795103": "Hi Lucas,\n\nSorry for the late reply, I forgot to answer before...\n\nThat's some very good points, they are all true.\nHowever, their relevance depends on the context and what we want to model.\n\nIn the case of GeoLifeCLEF 2022, we are interested in knowing where a species can be observed at some point of the year (without specifying it), e.g., all migrating species passing by, annual plants observable only at a time of the year, etc. That's why we don't give the time of the year.\nFor the \"time passing by\", in fact all the observations were collected in a single year (2019). Geographical shift of species distributions (due to extinctions or others) occur on a larger time scale and are thus not detectable here.\n\nIn any case, I don't think you will be able to recover part of this time information using the provided data."
  },
  "source": "meta"
}