{
  "id": 293654,
  "title": "Generating missing Target values",
  "url": "/competitions/g-research-crypto-forecasting/discussion/293654",
  "author_name": "",
  "post_date": "2021-12-06T11:49:25.548803600Z",
  "votes": 4,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hello everyone,</p>\n<p>Quick question on handling the missing Target values, just to clarify that I am not missing something important.</p>\n<p>The Target value in each time-step is a function of the Log(P[t+16]/P[t+1]), the weighted average of the Log(P[t+16]/P[t+1]) of the market for known and constant weights and beta, an arbitrary variable that is a function (fraction of rolling averages) of the above. Since no P data is missing, we can calculate the target values for each time-step (except the starting and last ones), right? Is there something that I am missing?</p>\n<p>The question might be dumb, but anyway thank you for taking the time to read this.</p>",
  "messages": [
    {
      "id": "1608445",
      "postDate": "12/06/2021 11:49:25",
      "content": "<p>Hello everyone,</p>\n<p>Quick question on handling the missing Target values, just to clarify that I am not missing something important.</p>\n<p>The Target value in each time-step is a function of the Log(P[t+16]/P[t+1]), the weighted average of the Log(P[t+16]/P[t+1]) of the market for known and constant weights and beta, an arbitrary variable that is a function (fraction of rolling averages) of the above. Since no P data is missing, we can calculate the target values for each time-step (except the starting and last ones), right? Is there something that I am missing?</p>\n<p>The question might be dumb, but anyway thank you for taking the time to read this.</p>",
      "rawMarkdown": "Hello everyone,\n\nQuick question on handling the missing Target values, just to clarify that I am not missing something important.\n\nThe Target value in each time-step is a function of the Log(P[t+16]/P[t+1]), the weighted average of the Log(P[t+16]/P[t+1]) of the market for known and constant weights and beta, an arbitrary variable that is a function (fraction of rolling averages) of the above. Since no P data is missing, we can calculate the target values for each time-step (except the starting and last ones), right? Is there something that I am missing?\n\nThe question might be dumb, but anyway thank you for taking the time to read this.",
      "votes": null
    },
    {
      "id": "1615814",
      "postDate": "12/12/2021 18:00:34",
      "content": "<p>Hi <a href=\"https://www.kaggle.com/fileleparalirima\" target=\"_blank\">@fileleparalirima</a>  I’ve concerned about this problem too and I wonder about the reason that why we got missing values in each coins?</p>\n<p>Anyone can tell me about this issue?<br>\nThank you</p>",
      "rawMarkdown": "Hi @fileleparalirima  I’ve concerned about this problem too and I wonder about the reason that why we got missing values in each coins?\n\nAnyone can tell me about this issue?\nThank you",
      "votes": null
    },
    {
      "id": "1621943",
      "postDate": "12/18/2021 08:17:42",
      "content": "<p>Take a look at the tutorial from g_research. There are missing values for certain timestamps i.e. not all assets have a value for every single minute. If you combine all the data into a dataframe indexed by the timestamp, you will see that some assets have NaN values for certain indices.</p>",
      "rawMarkdown": "Take a look at the tutorial from g_research. There are missing values for certain timestamps i.e. not all assets have a value for every single minute. If you combine all the data into a dataframe indexed by the timestamp, you will see that some assets have NaN values for certain indices.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1615814,
      "author_name": "premsurawut",
      "author_url": "",
      "post_date": "12/12/2021 18:00:34",
      "content": "<p>Hi <a href=\"https://www.kaggle.com/fileleparalirima\" target=\"_blank\">@fileleparalirima</a>  I’ve concerned about this problem too and I wonder about the reason that why we got missing values in each coins?</p>\n<p>Anyone can tell me about this issue?<br>\nThank you</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1621943,
      "author_name": "luchungi",
      "author_url": "",
      "post_date": "12/18/2021 08:17:42",
      "content": "<p>Take a look at the tutorial from g_research. There are missing values for certain timestamps i.e. not all assets have a value for every single minute. If you combine all the data into a dataframe indexed by the timestamp, you will see that some assets have NaN values for certain indices.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1608445": "Hello everyone,\n\nQuick question on handling the missing Target values, just to clarify that I am not missing something important.\n\nThe Target value in each time-step is a function of the Log(P[t+16]/P[t+1]), the weighted average of the Log(P[t+16]/P[t+1]) of the market for known and constant weights and beta, an arbitrary variable that is a function (fraction of rolling averages) of the above. Since no P data is missing, we can calculate the target values for each time-step (except the starting and last ones), right? Is there something that I am missing?\n\nThe question might be dumb, but anyway thank you for taking the time to read this.",
    "1615814": "Hi @fileleparalirima  I’ve concerned about this problem too and I wonder about the reason that why we got missing values in each coins?\n\nAnyone can tell me about this issue?\nThank you",
    "1621943": "Take a look at the tutorial from g_research. There are missing values for certain timestamps i.e. not all assets have a value for every single minute. If you combine all the data into a dataframe indexed by the timestamp, you will see that some assets have NaN values for certain indices."
  },
  "source": "meta"
}