{
  "id": 368538,
  "title": "3 dimensions Co-visitation matrix",
  "url": "/competitions/otto-recommender-system/discussion/368538",
  "author_name": "",
  "post_date": "2022-11-26T01:53:17.249968Z",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>In the previous public notebooks and discussions, the co-visitation matrix was created for two product pairs.<br>\nHowever, by creating a co-visitation matrix with three products and outputting product candidates for two products, the accuracy of the candidates may be improved if the conditions are met.</p>\n<p>Example: Co-visitation matrix 2D and 3D product linkage<br>\nCo-visitation matrix 2D : Product A → B<br>\nCo-visitation matrix 3D : Product A, B → C</p>\n<p>However, since it will be a large co-visitation matrix, memory and processing time will be a concern. In addition, since product candidates associated with two products are output during inference, there is a possibility that product candidates may not be available depending on the test data session.<br>\n*In particular, the 3D Co-visitation matrix cannot be used in a session where there is only one event in the test data.</p>\n<p>I would appreciate your comments if there are any errors in my comments.</p>",
  "messages": [
    {
      "id": "2043836",
      "postDate": "11/26/2022 01:53:17",
      "content": "<p>In the previous public notebooks and discussions, the co-visitation matrix was created for two product pairs.<br>\nHowever, by creating a co-visitation matrix with three products and outputting product candidates for two products, the accuracy of the candidates may be improved if the conditions are met.</p>\n<p>Example: Co-visitation matrix 2D and 3D product linkage<br>\nCo-visitation matrix 2D : Product A → B<br>\nCo-visitation matrix 3D : Product A, B → C</p>\n<p>However, since it will be a large co-visitation matrix, memory and processing time will be a concern. In addition, since product candidates associated with two products are output during inference, there is a possibility that product candidates may not be available depending on the test data session.<br>\n*In particular, the 3D Co-visitation matrix cannot be used in a session where there is only one event in the test data.</p>\n<p>I would appreciate your comments if there are any errors in my comments.</p>",
      "rawMarkdown": "In the previous public notebooks and discussions, the co-visitation matrix was created for two product pairs.\nHowever, by creating a co-visitation matrix with three products and outputting product candidates for two products, the accuracy of the candidates may be improved if the conditions are met.\n\nExample: Co-visitation matrix 2D and 3D product linkage\nCo-visitation matrix 2D : Product A → B\nCo-visitation matrix 3D : Product A, B → C\n\nHowever, since it will be a large co-visitation matrix, memory and processing time will be a concern. In addition, since product candidates associated with two products are output during inference, there is a possibility that product candidates may not be available depending on the test data session.\n*In particular, the 3D Co-visitation matrix cannot be used in a session where there is only one event in the test data.\n\nI would appreciate your comments if there are any errors in my comments.",
      "votes": null
    },
    {
      "id": "2079812",
      "postDate": "12/29/2022 16:14:06",
      "content": "<p>I think it is a very good idea.<br>\nHowever, I am not sure if it can actually be made, since it uses a huge amount of data.</p>",
      "rawMarkdown": "I think it is a very good idea.\nHowever, I am not sure if it can actually be made, since it uses a huge amount of data.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2079812,
      "author_name": "takuma0306",
      "author_url": "",
      "post_date": "12/29/2022 16:14:06",
      "content": "<p>I think it is a very good idea.<br>\nHowever, I am not sure if it can actually be made, since it uses a huge amount of data.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2043836": "In the previous public notebooks and discussions, the co-visitation matrix was created for two product pairs.\nHowever, by creating a co-visitation matrix with three products and outputting product candidates for two products, the accuracy of the candidates may be improved if the conditions are met.\n\nExample: Co-visitation matrix 2D and 3D product linkage\nCo-visitation matrix 2D : Product A → B\nCo-visitation matrix 3D : Product A, B → C\n\nHowever, since it will be a large co-visitation matrix, memory and processing time will be a concern. In addition, since product candidates associated with two products are output during inference, there is a possibility that product candidates may not be available depending on the test data session.\n*In particular, the 3D Co-visitation matrix cannot be used in a session where there is only one event in the test data.\n\nI would appreciate your comments if there are any errors in my comments.",
    "2079812": "I think it is a very good idea.\nHowever, I am not sure if it can actually be made, since it uses a huge amount of data."
  },
  "source": "meta"
}