{
  "id": 238318,
  "title": "Trying to Understand this Dataset",
  "url": "/competitions/indoor-location-navigation/discussion/238318",
  "author_name": "",
  "post_date": "2021-05-11T21:58:14.344899100Z",
  "votes": 3,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hello all,</p>\n<p>I recognize that this competition is just about over, I am not looking to figure a winning score, or even a good score, only to understand how to use the wifi data in order to run an ML pipeline in python.</p>\n<p>I am not well versed in coding, but have been making some strides with using scikit-learn. I've gone through the provided code and figured out how to map a specific trace on a specific floor in a specific building - but my questions are the following:</p>\n<p>1) Should I be trying to generate waypoints and interpolated wifi waypoints based on RSSI timestamps for each floor and each building in the training set? </p>\n<p>2) If I can manage to get all waypoint (x,y) data, wifi interpolated (x,y) and floor values, what is my target label? (Or am I not looking at this problem the correct way?)</p>\n<p>The general work I've done involves a set of data with a label, and then running ML concepts (Perceptron, Log Regression, MLP, SVR) but nothing as complicated as parsing data like this.</p>\n<p>I have tried analyzing and using the unified wifi datasets and have crashed my computer a few times, others just simply haven't finished computing after long runs</p>\n<p>I am genuinely lost and asking for basic guidance for trying to create and analyze a full training/testing dataset.</p>\n<p>I am sorry if these questions are not great, but I am trying to understand how to process all of this to the best of my abilities. </p>\n<p>Thank you.</p>",
  "messages": [
    {
      "id": "1303063",
      "postDate": "05/11/2021 21:58:14",
      "content": "<p>Hello all,</p>\n<p>I recognize that this competition is just about over, I am not looking to figure a winning score, or even a good score, only to understand how to use the wifi data in order to run an ML pipeline in python.</p>\n<p>I am not well versed in coding, but have been making some strides with using scikit-learn. I've gone through the provided code and figured out how to map a specific trace on a specific floor in a specific building - but my questions are the following:</p>\n<p>1) Should I be trying to generate waypoints and interpolated wifi waypoints based on RSSI timestamps for each floor and each building in the training set? </p>\n<p>2) If I can manage to get all waypoint (x,y) data, wifi interpolated (x,y) and floor values, what is my target label? (Or am I not looking at this problem the correct way?)</p>\n<p>The general work I've done involves a set of data with a label, and then running ML concepts (Perceptron, Log Regression, MLP, SVR) but nothing as complicated as parsing data like this.</p>\n<p>I have tried analyzing and using the unified wifi datasets and have crashed my computer a few times, others just simply haven't finished computing after long runs</p>\n<p>I am genuinely lost and asking for basic guidance for trying to create and analyze a full training/testing dataset.</p>\n<p>I am sorry if these questions are not great, but I am trying to understand how to process all of this to the best of my abilities. </p>\n<p>Thank you.</p>",
      "rawMarkdown": "Hello all,\n\nI recognize that this competition is just about over, I am not looking to figure a winning score, or even a good score, only to understand how to use the wifi data in order to run an ML pipeline in python.\n\nI am not well versed in coding, but have been making some strides with using scikit-learn. I've gone through the provided code and figured out how to map a specific trace on a specific floor in a specific building - but my questions are the following:\n\n1) Should I be trying to generate waypoints and interpolated wifi waypoints based on RSSI timestamps for each floor and each building in the training set? \n\n2) If I can manage to get all waypoint (x,y) data, wifi interpolated (x,y) and floor values, what is my target label? (Or am I not looking at this problem the correct way?)\n\nThe general work I've done involves a set of data with a label, and then running ML concepts (Perceptron, Log Regression, MLP, SVR) but nothing as complicated as parsing data like this.\n\nI have tried analyzing and using the unified wifi datasets and have crashed my computer a few times, others just simply haven't finished computing after long runs\n\nI am genuinely lost and asking for basic guidance for trying to create and analyze a full training/testing dataset.\n\nI am sorry if these questions are not great, but I am trying to understand how to process all of this to the best of my abilities. \n\nThank you.",
      "votes": null
    },
    {
      "id": "1303243",
      "postDate": "05/12/2021 01:26:19",
      "content": "<p>Take a look at sample_submission.csv - those rows are exactly what you're trying to predict (the values for those sites/paths/timestamps). You have to fill in the x, y, and floor values for those \"waypoints\".  Many of those waypoints (x,y values) will be present at least once in the training data, but some will not be, and you'll have to figure out how to get values for those too.</p>\n<p>Since there is a lot of data in the full set, I'd suggest getting started with just a single site first. That will let you iterate on just a small subset of the data, and get your initial model up and running.</p>",
      "rawMarkdown": "Take a look at sample_submission.csv - those rows are exactly what you're trying to predict (the values for those sites/paths/timestamps). You have to fill in the x, y, and floor values for those \"waypoints\".  Many of those waypoints (x,y values) will be present at least once in the training data, but some will not be, and you'll have to figure out how to get values for those too.\n\nSince there is a lot of data in the full set, I'd suggest getting started with just a single site first. That will let you iterate on just a small subset of the data, and get your initial model up and running.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1303243,
      "author_name": "chris62",
      "author_url": "",
      "post_date": "05/12/2021 01:26:19",
      "content": "<p>Take a look at sample_submission.csv - those rows are exactly what you're trying to predict (the values for those sites/paths/timestamps). You have to fill in the x, y, and floor values for those \"waypoints\".  Many of those waypoints (x,y values) will be present at least once in the training data, but some will not be, and you'll have to figure out how to get values for those too.</p>\n<p>Since there is a lot of data in the full set, I'd suggest getting started with just a single site first. That will let you iterate on just a small subset of the data, and get your initial model up and running.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1303063": "Hello all,\n\nI recognize that this competition is just about over, I am not looking to figure a winning score, or even a good score, only to understand how to use the wifi data in order to run an ML pipeline in python.\n\nI am not well versed in coding, but have been making some strides with using scikit-learn. I've gone through the provided code and figured out how to map a specific trace on a specific floor in a specific building - but my questions are the following:\n\n1) Should I be trying to generate waypoints and interpolated wifi waypoints based on RSSI timestamps for each floor and each building in the training set? \n\n2) If I can manage to get all waypoint (x,y) data, wifi interpolated (x,y) and floor values, what is my target label? (Or am I not looking at this problem the correct way?)\n\nThe general work I've done involves a set of data with a label, and then running ML concepts (Perceptron, Log Regression, MLP, SVR) but nothing as complicated as parsing data like this.\n\nI have tried analyzing and using the unified wifi datasets and have crashed my computer a few times, others just simply haven't finished computing after long runs\n\nI am genuinely lost and asking for basic guidance for trying to create and analyze a full training/testing dataset.\n\nI am sorry if these questions are not great, but I am trying to understand how to process all of this to the best of my abilities. \n\nThank you.",
    "1303243": "Take a look at sample_submission.csv - those rows are exactly what you're trying to predict (the values for those sites/paths/timestamps). You have to fill in the x, y, and floor values for those \"waypoints\".  Many of those waypoints (x,y values) will be present at least once in the training data, but some will not be, and you'll have to figure out how to get values for those too.\n\nSince there is a lot of data in the full set, I'd suggest getting started with just a single site first. That will let you iterate on just a small subset of the data, and get your initial model up and running."
  },
  "source": "meta"
}