{
  "id": 81755,
  "title": "Confusion matrix and test accuracy",
  "url": "/competitions/human-protein-atlas-image-classification/discussion/81755",
  "author_name": "",
  "post_date": "2019-02-24T17:23:06.081002100Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>let's say we have test set true labels, how can we calculate test accuracy and confusion matrix? Is there any sample code to learn from it?</p>",
  "messages": [
    {
      "id": "477485",
      "postDate": "02/24/2019 17:23:06",
      "content": "<p>let's say we have test set true labels, how can we calculate test accuracy and confusion matrix? Is there any sample code to learn from it?</p>",
      "rawMarkdown": "let's say we have test set true labels, how can we calculate test accuracy and confusion matrix? Is there any sample code to learn from it?",
      "votes": null
    },
    {
      "id": "477580",
      "postDate": "02/24/2019 22:29:21",
      "content": "<p>Have a look at the baseline kernels. As this is a multi label detection with no bounding boxes, it is a bit hard to say which label was mistaken for another if you have two. You can possibly gain some insight if you have only one wrong label (one missing (fn) one extra (fp)) but i doubt it will count for much. </p>",
      "rawMarkdown": "Have a look at the baseline kernels. As this is a multi label detection with no bounding boxes, it is a bit hard to say which label was mistaken for another if you have two. You can possibly gain some insight if you have only one wrong label (one missing (fn) one extra (fp)) but i doubt it will count for much.",
      "votes": null
    },
    {
      "id": "478126",
      "postDate": "02/25/2019 19:15:15",
      "content": "<p>It's suggestable to perform cross-validation first with a proper test-train split. This will give you a fair idea about the confusion matrix and how better your model will perform.</p>",
      "rawMarkdown": "It's suggestable to perform cross-validation first with a proper test-train split. This will give you a fair idea about the confusion matrix and how better your model will perform.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 477580,
      "author_name": "moshel",
      "author_url": "",
      "post_date": "02/24/2019 22:29:21",
      "content": "<p>Have a look at the baseline kernels. As this is a multi label detection with no bounding boxes, it is a bit hard to say which label was mistaken for another if you have two. You can possibly gain some insight if you have only one wrong label (one missing (fn) one extra (fp)) but i doubt it will count for much. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 478126,
      "author_name": "maheshtripathi",
      "author_url": "",
      "post_date": "02/25/2019 19:15:15",
      "content": "<p>It's suggestable to perform cross-validation first with a proper test-train split. This will give you a fair idea about the confusion matrix and how better your model will perform.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "477485": "let's say we have test set true labels, how can we calculate test accuracy and confusion matrix? Is there any sample code to learn from it?",
    "477580": "Have a look at the baseline kernels. As this is a multi label detection with no bounding boxes, it is a bit hard to say which label was mistaken for another if you have two. You can possibly gain some insight if you have only one wrong label (one missing (fn) one extra (fp)) but i doubt it will count for much.",
    "478126": "It's suggestable to perform cross-validation first with a proper test-train split. This will give you a fair idea about the confusion matrix and how better your model will perform."
  },
  "source": "meta"
}