{
  "id": 22282,
  "title": "Gap in the sales of the users",
  "url": "/competitions/grupo-bimbo-inventory-demand/discussion/22282",
  "author_name": "",
  "post_date": "2016-07-14T23:05:11.470Z",
  "votes": null,
  "comment_count": 3,
  "views": 445,
  "content": "<p>I am using PostgreSQL to deal with the data and I perform the following sql query to know the information of the cliente_id = 26 and producto_id = 31518.</p>\n\n<p>SELECT * \nFROM train\nWHERE cliente_id = 26 AND producto_id = 31518\nORDER BY producto_id, semana, agencia_id  ; </p>\n\n<p>The result is </p>\n\n<h2>semana       agencia_id   canal_id     ruta_sak     cliente_id   producto_id  venta_uni_hoy   venta_hoy       dev_uni_proxima  dev_proxima     demanda_uni_equil  </h2>\n\n<p>4            2655         2            4189         26           31518        10              305.29999       0                0               10       </p>\n\n<p>In the test data the following information of this user and product is given: </p>\n\n<h2>id           semana       agencia_id   canal_id     ruta_sak     cliente_id   producto_id  </h2>\n\n<p>1569352      10           2655         2            4189         26           31518        </p>\n\n<p>I noticed there is a gap between the week  4  and  week 10. The data is very noisy and tricky. How can I model this behaviour in the data. I think the simplest model is to forecast in the test data with the same information in the train data. </p>",
  "messages": [
    {
      "id": "127685",
      "postDate": "07/14/2016 23:05:11",
      "content": "<p>I am using PostgreSQL to deal with the data and I perform the following sql query to know the information of the cliente_id = 26 and producto_id = 31518.</p>\n\n<p>SELECT * \nFROM train\nWHERE cliente_id = 26 AND producto_id = 31518\nORDER BY producto_id, semana, agencia_id  ; </p>\n\n<p>The result is </p>\n\n<h2>semana       agencia_id   canal_id     ruta_sak     cliente_id   producto_id  venta_uni_hoy   venta_hoy       dev_uni_proxima  dev_proxima     demanda_uni_equil  </h2>\n\n<p>4            2655         2            4189         26           31518        10              305.29999       0                0               10       </p>\n\n<p>In the test data the following information of this user and product is given: </p>\n\n<h2>id           semana       agencia_id   canal_id     ruta_sak     cliente_id   producto_id  </h2>\n\n<p>1569352      10           2655         2            4189         26           31518        </p>\n\n<p>I noticed there is a gap between the week  4  and  week 10. The data is very noisy and tricky. How can I model this behaviour in the data. I think the simplest model is to forecast in the test data with the same information in the train data. </p>",
      "rawMarkdown": "I am using PostgreSQL to deal with the data and I perform the following sql query to know the information of the cliente_id = 26 and producto_id = 31518.\r\n\r\nSELECT * \r\nFROM train\r\nWHERE cliente_id = 26 AND producto_id = 31518\r\nORDER BY producto_id, semana, agencia_id  ; \r\n\r\nThe result is \r\n\r\nsemana       agencia_id   canal_id     ruta_sak     cliente_id   producto_id  venta_uni_hoy   venta_hoy       dev_uni_proxima  dev_proxima     demanda_uni_equil  \r\n------------------------------------------------------------------------------------------------------------------------------------------------------------------\r\n4            2655         2            4189         26           31518        10              305.29999       0                0               10       \r\n\r\n\r\n\r\nIn the test data the following information of this user and product is given: \r\n\r\n\r\nid           semana       agencia_id   canal_id     ruta_sak     cliente_id   producto_id  \r\n-------------------------------------------------------------------------------------------\r\n1569352      10           2655         2            4189         26           31518        \r\n\r\n\r\n\r\nI noticed there is a gap between the week  4  and  week 10. The data is very noisy and tricky. How can I model this behaviour in the data. I think the simplest model is to forecast in the test data with the same information in the train data.",
      "votes": null
    },
    {
      "id": "127823",
      "postDate": "07/15/2016 15:12:17",
      "content": "<p>I think the fact that the client has only information of the week 4 in the train data is due to the client did not buy in the other weeks. </p>",
      "rawMarkdown": "I think the fact that the client has only information of the week 4 in the train data is due to the client did not buy in the other weeks.",
      "votes": null
    },
    {
      "id": "127826",
      "postDate": "07/15/2016 15:47:15",
      "content": "<p>[quote=Ricardo Chess;127823]</p>\n\n<p>I think the fact that the client has only information of the week 4 in the train data is due to the client did not buy in the other weeks. </p>\n\n<p>[/quote]</p>\n\n<p>You are right. It is explained in detail in other threads of this competition, but the short answer is that we have only records for those tuples in which either the demand or the returns were &gt; 0. Hence, if there is no record for a particular combination, you can assume that both demand and returns were 0.</p>",
      "rawMarkdown": "[quote=Ricardo Chess;127823]\r\n\r\nI think the fact that the client has only information of the week 4 in the train data is due to the client did not buy in the other weeks. \r\n\r\n[/quote]\r\n\r\nYou are right. It is explained in detail in other threads of this competition, but the short answer is that we have only records for those tuples in which either the demand or the returns were > 0. Hence, if there is no record for a particular combination, you can assume that both demand and returns were 0.",
      "votes": null
    },
    {
      "id": "127833",
      "postDate": "07/15/2016 16:31:30",
      "content": "<p>Thanks a lot for your answer. </p>",
      "rawMarkdown": "Thanks a lot for your answer.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 127823,
      "author_name": "ricardorios",
      "author_url": "",
      "post_date": "07/15/2016 15:12:17",
      "content": "<p>I think the fact that the client has only information of the week 4 in the train data is due to the client did not buy in the other weeks. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 127826,
      "author_name": "carrdelling",
      "author_url": "",
      "post_date": "07/15/2016 15:47:15",
      "content": "<p>[quote=Ricardo Chess;127823]</p>\n\n<p>I think the fact that the client has only information of the week 4 in the train data is due to the client did not buy in the other weeks. </p>\n\n<p>[/quote]</p>\n\n<p>You are right. It is explained in detail in other threads of this competition, but the short answer is that we have only records for those tuples in which either the demand or the returns were &gt; 0. Hence, if there is no record for a particular combination, you can assume that both demand and returns were 0.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 127833,
      "author_name": "ricardorios",
      "author_url": "",
      "post_date": "07/15/2016 16:31:30",
      "content": "<p>Thanks a lot for your answer. </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "127685": "I am using PostgreSQL to deal with the data and I perform the following sql query to know the information of the cliente_id = 26 and producto_id = 31518.\r\n\r\nSELECT * \r\nFROM train\r\nWHERE cliente_id = 26 AND producto_id = 31518\r\nORDER BY producto_id, semana, agencia_id  ; \r\n\r\nThe result is \r\n\r\nsemana       agencia_id   canal_id     ruta_sak     cliente_id   producto_id  venta_uni_hoy   venta_hoy       dev_uni_proxima  dev_proxima     demanda_uni_equil  \r\n------------------------------------------------------------------------------------------------------------------------------------------------------------------\r\n4            2655         2            4189         26           31518        10              305.29999       0                0               10       \r\n\r\n\r\n\r\nIn the test data the following information of this user and product is given: \r\n\r\n\r\nid           semana       agencia_id   canal_id     ruta_sak     cliente_id   producto_id  \r\n-------------------------------------------------------------------------------------------\r\n1569352      10           2655         2            4189         26           31518        \r\n\r\n\r\n\r\nI noticed there is a gap between the week  4  and  week 10. The data is very noisy and tricky. How can I model this behaviour in the data. I think the simplest model is to forecast in the test data with the same information in the train data.",
    "127823": "I think the fact that the client has only information of the week 4 in the train data is due to the client did not buy in the other weeks.",
    "127826": "[quote=Ricardo Chess;127823]\r\n\r\nI think the fact that the client has only information of the week 4 in the train data is due to the client did not buy in the other weeks. \r\n\r\n[/quote]\r\n\r\nYou are right. It is explained in detail in other threads of this competition, but the short answer is that we have only records for those tuples in which either the demand or the returns were > 0. Hence, if there is no record for a particular combination, you can assume that both demand and returns were 0.",
    "127833": "Thanks a lot for your answer."
  },
  "source": "meta"
}