{
  "id": 315735,
  "title": "Profiling customers",
  "url": "/competitions/h-and-m-personalized-fashion-recommendations/discussion/315735",
  "author_name": "Eduardo Catarino",
  "post_date": "2022-03-29T14:07:58.266000",
  "votes": 1,
  "comment_count": 0,
  "views": 0,
  "content": "<p>As a first step towards a good recomendations seems to me try to profile the type of customer I'm trying to provide information.</p>\n<p>For that I merged the 3 datasets into one and then, reduced the information to a smaller data set:</p>\n<p>| customerId | Age  | Mean of sales channel | Number of visits to H&amp;M | Mean index_group_no |</p>\n<p>customerId - Customer ID<br>\nAge - Customer's Age<br>\nMean of sales channel - Minimum value 1  means only on store purchases  - Maximum value 2 means only only purchages<br>\nindex_group_no    index_group_name<br>\n    1   Ladieswear<br>\n    4   Baby/Children<br>\n    3   Menswear<br>\n    5   Sport<br>\n    2   Divided</p>\n<p>After this pre-processing step I'm using MeanShift clustering algorithm to try to get labels for each customer and get a first profile to him/her.</p>\n<p>After this step I will try to diferentiate users by the profiles results and refine my next steps.</p>\n<p>Does anyone tryed this? Ideas are appreciated.</p>",
  "messages": [
    {
      "id": 1738765,
      "postDate": "2022-03-29T14:07:58.267Z",
      "content": "<p>As a first step towards a good recomendations seems to me try to profile the type of customer I'm trying to provide information.</p>\n<p>For that I merged the 3 datasets into one and then, reduced the information to a smaller data set:</p>\n<p>| customerId | Age  | Mean of sales channel | Number of visits to H&amp;M | Mean index_group_no |</p>\n<p>customerId - Customer ID<br>\nAge - Customer's Age<br>\nMean of sales channel - Minimum value 1  means only on store purchases  - Maximum value 2 means only only purchages<br>\nindex_group_no    index_group_name<br>\n    1   Ladieswear<br>\n    4   Baby/Children<br>\n    3   Menswear<br>\n    5   Sport<br>\n    2   Divided</p>\n<p>After this pre-processing step I'm using MeanShift clustering algorithm to try to get labels for each customer and get a first profile to him/her.</p>\n<p>After this step I will try to diferentiate users by the profiles results and refine my next steps.</p>\n<p>Does anyone tryed this? Ideas are appreciated.</p>",
      "rawMarkdown": "As a first step towards a good recomendations seems to me try to profile the type of customer I'm trying to provide information.\n\nFor that I merged the 3 datasets into one and then, reduced the information to a smaller data set:\n\n| customerId | Age  | Mean of sales channel | Number of visits to H&M | Mean index_group_no |\n\ncustomerId - Customer ID\nAge - Customer's Age\nMean of sales channel - Minimum value 1  means only on store purchases  - Maximum value 2 means only only purchages\nindex_group_no\tindex_group_name\n\t1\tLadieswear\n\t4\tBaby/Children\n\t3\tMenswear\n\t5\tSport\n\t2\tDivided\n\n\nAfter this pre-processing step I'm using MeanShift clustering algorithm to try to get labels for each customer and get a first profile to him/her.\n\nAfter this step I will try to diferentiate users by the profiles results and refine my next steps.\n\nDoes anyone tryed this? Ideas are appreciated.",
      "votes": 1
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1738765": "As a first step towards a good recomendations seems to me try to profile the type of customer I'm trying to provide information.\n\nFor that I merged the 3 datasets into one and then, reduced the information to a smaller data set:\n\n| customerId | Age  | Mean of sales channel | Number of visits to H&M | Mean index_group_no |\n\ncustomerId - Customer ID\nAge - Customer's Age\nMean of sales channel - Minimum value 1  means only on store purchases  - Maximum value 2 means only only purchages\nindex_group_no\tindex_group_name\n\t1\tLadieswear\n\t4\tBaby/Children\n\t3\tMenswear\n\t5\tSport\n\t2\tDivided\n\n\nAfter this pre-processing step I'm using MeanShift clustering algorithm to try to get labels for each customer and get a first profile to him/her.\n\nAfter this step I will try to diferentiate users by the profiles results and refine my next steps.\n\nDoes anyone tryed this? Ideas are appreciated."
  }
}