{
  "id": 364223,
  "title": "How to incorporate event-type with DL models/embeddings?",
  "url": "/competitions/otto-recommender-system/discussion/364223",
  "author_name": "",
  "post_date": "2022-11-05T09:24:11.808164800Z",
  "votes": 7,
  "comment_count": 1,
  "views": 0,
  "content": "<p>What are the best approaches to combing the event-type (click/basket/etc) with the item if, when learning sequence/TS features for sessions? </p>\n<p>From what I saw in the literature, there's item/user/session specific features, but I didn't see anything decent for the context of the item itself , beyond sequences of items, sequences of sessions and all-all attention (ineffecient) or attention within each sequence/per-\"column\" (which misses context). </p>\n<p>There's other tricks I can think of, but I'm sure it's been done somewhere?</p>",
  "messages": [
    {
      "id": "2017977",
      "postDate": "11/05/2022 09:24:11",
      "content": "<p>What are the best approaches to combing the event-type (click/basket/etc) with the item if, when learning sequence/TS features for sessions? </p>\n<p>From what I saw in the literature, there's item/user/session specific features, but I didn't see anything decent for the context of the item itself , beyond sequences of items, sequences of sessions and all-all attention (ineffecient) or attention within each sequence/per-\"column\" (which misses context). </p>\n<p>There's other tricks I can think of, but I'm sure it's been done somewhere?</p>",
      "rawMarkdown": "What are the best approaches to combing the event-type (click/basket/etc) with the item if, when learning sequence/TS features for sessions? \n\nFrom what I saw in the literature, there's item/user/session specific features, but I didn't see anything decent for the context of the item itself , beyond sequences of items, sequences of sessions and all-all attention (ineffecient) or attention within each sequence/per-\"column\" (which misses context). \n\nThere's other tricks I can think of, but I'm sure it's been done somewhere?",
      "votes": null
    },
    {
      "id": "2053449",
      "postDate": "12/03/2022 09:45:41",
      "content": "<p>I watched some matrix factorization courses and they were using ratings/scores as targets. Event type is ordinal and similar to ratings in that sense.</p>",
      "rawMarkdown": "I watched some matrix factorization courses and they were using ratings/scores as targets. Event type is ordinal and similar to ratings in that sense.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2053449,
      "author_name": "gunesevitan",
      "author_url": "",
      "post_date": "12/03/2022 09:45:41",
      "content": "<p>I watched some matrix factorization courses and they were using ratings/scores as targets. Event type is ordinal and similar to ratings in that sense.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2017977": "What are the best approaches to combing the event-type (click/basket/etc) with the item if, when learning sequence/TS features for sessions? \n\nFrom what I saw in the literature, there's item/user/session specific features, but I didn't see anything decent for the context of the item itself , beyond sequences of items, sequences of sessions and all-all attention (ineffecient) or attention within each sequence/per-\"column\" (which misses context). \n\nThere's other tricks I can think of, but I'm sure it's been done somewhere?",
    "2053449": "I watched some matrix factorization courses and they were using ratings/scores as targets. Event type is ordinal and similar to ratings in that sense."
  },
  "source": "meta"
}