{
  "id": 356852,
  "title": "Feature engineering ideas",
  "url": "/competitions/tabular-playground-series-oct-2022/discussion/356852",
  "author_name": "",
  "post_date": "2022-10-02T07:47:27.075186400Z",
  "votes": 27,
  "comment_count": 5,
  "views": 0,
  "content": "<h3>New features</h3>\n<p>There is lots of opportunity for feature engineering this competition. We have x, y, z coordinates for position and velocity. This means we can compute the <strong>Euclidean</strong>, <strong>Manhattan</strong> or <strong>Minkowski distance</strong> to the goal. </p>\n<p><a href=\"https://postimg.cc/bSSbDDpj\" target=\"_blank\"><img src=\"https://i.postimg.cc/J0PQFZFt/1-v-At-QZb-ROu-Tdp36a-QQ8cq-BA.png\" alt=\"1-v-At-QZb-ROu-Tdp36a-QQ8cq-BA.png\"></a></p>\n<p>We also have the <strong>velocity</strong>, so another idea could be to compute the <strong>absolute speed</strong> of the ball / cars using the formula speed = sqrt(v_x^2+v_y^2+v_z^2).</p>\n<p>We have multiple cars so how about computing the distances between themselves, or the <strong>average/maximum/minimum distance</strong> of the 3 cars on one team to the goal. </p>\n<p>There are two teams so we can compute the average/maximum <strong>distance between the two teams</strong>. How many opposition players are there <strong>between the ball and the goal</strong>? </p>\n<p>Is there a 'goal keeper', i.e. a player defending in goal, or is it an <strong>open goal</strong>? The right feature for this would be the <strong>minimum distance of a player to the goal</strong>. </p>\n<p>Depending on the ability of the players, being <strong>elevated off the ground</strong> can often signify a player is taking/defending a shot. So the z-coordinates could be quite useful.</p>\n<h3>Enjoy some gifs</h3>\n<p><a href=\"https://postimg.cc/bGFKHJKv\" target=\"_blank\"><img src=\"https://i.postimg.cc/gJYmpLs3/giphy-car1.gif\" alt=\"giphy-car1.gif\"></a></p>\n<p><br><br>\n<br><br>\n<br></p>\n<p><a href=\"https://postimg.cc/bZwCCP0S\" target=\"_blank\"><img src=\"https://i.postimg.cc/Qdp2MxHm/200-car2.gif\" alt=\"200-car2.gif\"></a></p>",
  "messages": [
    {
      "id": "1966875",
      "postDate": "10/02/2022 07:47:27",
      "content": "<h3>New features</h3>\n<p>There is lots of opportunity for feature engineering this competition. We have x, y, z coordinates for position and velocity. This means we can compute the <strong>Euclidean</strong>, <strong>Manhattan</strong> or <strong>Minkowski distance</strong> to the goal. </p>\n<p><a href=\"https://postimg.cc/bSSbDDpj\" target=\"_blank\"><img src=\"https://i.postimg.cc/J0PQFZFt/1-v-At-QZb-ROu-Tdp36a-QQ8cq-BA.png\" alt=\"1-v-At-QZb-ROu-Tdp36a-QQ8cq-BA.png\"></a></p>\n<p>We also have the <strong>velocity</strong>, so another idea could be to compute the <strong>absolute speed</strong> of the ball / cars using the formula speed = sqrt(v_x^2+v_y^2+v_z^2).</p>\n<p>We have multiple cars so how about computing the distances between themselves, or the <strong>average/maximum/minimum distance</strong> of the 3 cars on one team to the goal. </p>\n<p>There are two teams so we can compute the average/maximum <strong>distance between the two teams</strong>. How many opposition players are there <strong>between the ball and the goal</strong>? </p>\n<p>Is there a 'goal keeper', i.e. a player defending in goal, or is it an <strong>open goal</strong>? The right feature for this would be the <strong>minimum distance of a player to the goal</strong>. </p>\n<p>Depending on the ability of the players, being <strong>elevated off the ground</strong> can often signify a player is taking/defending a shot. So the z-coordinates could be quite useful.</p>\n<h3>Enjoy some gifs</h3>\n<p><a href=\"https://postimg.cc/bGFKHJKv\" target=\"_blank\"><img src=\"https://i.postimg.cc/gJYmpLs3/giphy-car1.gif\" alt=\"giphy-car1.gif\"></a></p>\n<p><br><br>\n<br><br>\n<br></p>\n<p><a href=\"https://postimg.cc/bZwCCP0S\" target=\"_blank\"><img src=\"https://i.postimg.cc/Qdp2MxHm/200-car2.gif\" alt=\"200-car2.gif\"></a></p>",
      "rawMarkdown": "### New features\n\nThere is lots of opportunity for feature engineering this competition. We have x, y, z coordinates for position and velocity. This means we can compute the **Euclidean**, **Manhattan** or **Minkowski distance** to the goal. \n\n[![1-v-At-QZb-ROu-Tdp36a-QQ8cq-BA.png](https://i.postimg.cc/J0PQFZFt/1-v-At-QZb-ROu-Tdp36a-QQ8cq-BA.png)](https://postimg.cc/bSSbDDpj)\n\nWe also have the **velocity**, so another idea could be to compute the **absolute speed** of the ball / cars using the formula speed = sqrt(v_x^2+v_y^2+v_z^2).\n\nWe have multiple cars so how about computing the distances between themselves, or the **average/maximum/minimum distance** of the 3 cars on one team to the goal. \n\nThere are two teams so we can compute the average/maximum **distance between the two teams**. How many opposition players are there **between the ball and the goal**? \n\nIs there a 'goal keeper', i.e. a player defending in goal, or is it an **open goal**? The right feature for this would be the **minimum distance of a player to the goal**. \n\nDepending on the ability of the players, being **elevated off the ground** can often signify a player is taking/defending a shot. So the z-coordinates could be quite useful.\n\n### Enjoy some gifs\n\n[![giphy-car1.gif](https://i.postimg.cc/gJYmpLs3/giphy-car1.gif)](https://postimg.cc/bGFKHJKv)\n\n<br>\n<br>\n<br>\n\n[![200-car2.gif](https://i.postimg.cc/Qdp2MxHm/200-car2.gif)](https://postimg.cc/bZwCCP0S)",
      "votes": null
    },
    {
      "id": "1968039",
      "postDate": "10/02/2022 20:31:35",
      "content": "<p>Interesting ideas for feature engineering. Those gifs make me wanna play Rocket League now lol</p>",
      "rawMarkdown": "Interesting ideas for feature engineering. Those gifs make me wanna play Rocket League now lol",
      "votes": null
    },
    {
      "id": "1971027",
      "postDate": "10/04/2022 12:23:41",
      "content": "<p>interesting ideas for feature engineering!</p>",
      "rawMarkdown": "interesting ideas for feature engineering!",
      "votes": null
    },
    {
      "id": "1972595",
      "postDate": "10/05/2022 08:29:38",
      "content": "<p>More about Distance Measures<br>\n<a href=\"https://postimg.cc/gwqJzpG3\" target=\"_blank\"><img src=\"https://i.postimg.cc/g016Kc5t/Screenshot-2022-10-05-142726.png\" alt=\"Screenshot-2022-10-05-142726.png\"></a></p>",
      "rawMarkdown": "More about Distance Measures\n[![Screenshot-2022-10-05-142726.png](https://i.postimg.cc/g016Kc5t/Screenshot-2022-10-05-142726.png)](https://postimg.cc/gwqJzpG3)",
      "votes": null
    },
    {
      "id": "2007494",
      "postDate": "10/28/2022 09:14:05",
      "content": "<p><a href=\"https://www.kaggle.com/samuelcortinhas\" target=\"_blank\">@samuelcortinhas</a> How to calculate distance b/w ball and goal. For ball we have the coordinates but what are the coordinates for both the goals ?</p>",
      "rawMarkdown": "samuelcortinhas How to calculate distance b/w ball and goal. For ball we have the coordinates but what are the coordinates for both the goals ?",
      "votes": null
    },
    {
      "id": "2007537",
      "postDate": "10/28/2022 09:43:07",
      "content": "<p>You can estimate this from the training data. If I remember off the top of my head correctly it is something like (0,+/-102.2,1.6)</p>",
      "rawMarkdown": "You can estimate this from the training data. If I remember off the top of my head correctly it is something like (0,+/-102.2,1.6)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1968039,
      "author_name": "chazzer",
      "author_url": "",
      "post_date": "10/02/2022 20:31:35",
      "content": "<p>Interesting ideas for feature engineering. Those gifs make me wanna play Rocket League now lol</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1971027,
      "author_name": "saipranoy",
      "author_url": "",
      "post_date": "10/04/2022 12:23:41",
      "content": "<p>interesting ideas for feature engineering!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1972595,
      "author_name": "gazu468",
      "author_url": "",
      "post_date": "10/05/2022 08:29:38",
      "content": "<p>More about Distance Measures<br>\n<a href=\"https://postimg.cc/gwqJzpG3\" target=\"_blank\"><img src=\"https://i.postimg.cc/g016Kc5t/Screenshot-2022-10-05-142726.png\" alt=\"Screenshot-2022-10-05-142726.png\"></a></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 2007494,
      "author_name": "frostpiee",
      "author_url": "",
      "post_date": "10/28/2022 09:14:05",
      "content": "<p><a href=\"https://www.kaggle.com/samuelcortinhas\" target=\"_blank\">@samuelcortinhas</a> How to calculate distance b/w ball and goal. For ball we have the coordinates but what are the coordinates for both the goals ?</p>",
      "votes": null,
      "replies": [
        {
          "id": 2007537,
          "author_name": "samuelcortinhas",
          "author_url": "",
          "post_date": "10/28/2022 09:43:07",
          "content": "<p>You can estimate this from the training data. If I remember off the top of my head correctly it is something like (0,+/-102.2,1.6)</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1966875": "### New features\n\nThere is lots of opportunity for feature engineering this competition. We have x, y, z coordinates for position and velocity. This means we can compute the **Euclidean**, **Manhattan** or **Minkowski distance** to the goal. \n\n[![1-v-At-QZb-ROu-Tdp36a-QQ8cq-BA.png](https://i.postimg.cc/J0PQFZFt/1-v-At-QZb-ROu-Tdp36a-QQ8cq-BA.png)](https://postimg.cc/bSSbDDpj)\n\nWe also have the **velocity**, so another idea could be to compute the **absolute speed** of the ball / cars using the formula speed = sqrt(v_x^2+v_y^2+v_z^2).\n\nWe have multiple cars so how about computing the distances between themselves, or the **average/maximum/minimum distance** of the 3 cars on one team to the goal. \n\nThere are two teams so we can compute the average/maximum **distance between the two teams**. How many opposition players are there **between the ball and the goal**? \n\nIs there a 'goal keeper', i.e. a player defending in goal, or is it an **open goal**? The right feature for this would be the **minimum distance of a player to the goal**. \n\nDepending on the ability of the players, being **elevated off the ground** can often signify a player is taking/defending a shot. So the z-coordinates could be quite useful.\n\n### Enjoy some gifs\n\n[![giphy-car1.gif](https://i.postimg.cc/gJYmpLs3/giphy-car1.gif)](https://postimg.cc/bGFKHJKv)\n\n<br>\n<br>\n<br>\n\n[![200-car2.gif](https://i.postimg.cc/Qdp2MxHm/200-car2.gif)](https://postimg.cc/bZwCCP0S)",
    "1968039": "Interesting ideas for feature engineering. Those gifs make me wanna play Rocket League now lol",
    "1971027": "interesting ideas for feature engineering!",
    "1972595": "More about Distance Measures\n[![Screenshot-2022-10-05-142726.png](https://i.postimg.cc/g016Kc5t/Screenshot-2022-10-05-142726.png)](https://postimg.cc/gwqJzpG3)",
    "2007494": "samuelcortinhas How to calculate distance b/w ball and goal. For ball we have the coordinates but what are the coordinates for both the goals ?",
    "2007537": "You can estimate this from the training data. If I remember off the top of my head correctly it is something like (0,+/-102.2,1.6)"
  },
  "source": "meta"
}