I'm not sure if I'm missing something because no one else seems to be mentioning this, but I'm finding duplicates in the training data. For the same Semana, Cliente_ID, and Producto_ID I see two (at least, there may be more) records with different Demanda_uni_equil. For example row 1522 and 134598.
", "messages": [ { "id": "124475", "postDate": "06/19/2016 02:47:53", "content": "I'm not sure if I'm missing something because no one else seems to be mentioning this, but I'm finding duplicates in the training data. For the same Semana, Cliente_ID, and Producto_ID I see two (at least, there may be more) records with different Demanda_uni_equil. For example row 1522 and 134598.
", "rawMarkdown": "I'm not sure if I'm missing something because no one else seems to be mentioning this, but I'm finding duplicates in the training data. For the same Semana, Cliente_ID, and Producto_ID I see two (at least, there may be more) records with different Demanda_uni_equil. For example row 1522 and 134598.", "votes": null }, { "id": "124476", "postDate": "06/19/2016 02:50:24", "content": "Same client may be taking same product from different Agencies... Can you check it out that they have different agency ID. Actually everyone already knows it. That's why there is agency address files so that we can visualize location of agency, clients and routes connecting them. Basically what you can do is take agencies as nodes, and their common Route as edges. Their common client should lie on that route. { Just a hypothesis..}. Also i think there should be negative correlation between demand for common products in such agencies (i.e. two agencies serving on same route , supplying same product to same client)
", "rawMarkdown": "Same client may be taking same product from different Agencies... Can you check it out that they have different agency ID. Actually everyone already knows it. That's why there is agency address files so that we can visualize location of agency, clients and routes connecting them. Basically what you can do is take agencies as nodes, and their common Route as edges. Their common client should lie on that route. { Just a hypothesis..}. Also i think there should be negative correlation between demand for common products in such agencies (i.e. two agencies serving on same route , supplying same product to same client)", "votes": null }, { "id": "124477", "postDate": "06/19/2016 03:57:05", "content": "[quote=Shahnawaz Akhtar;124476]
\n\nSame client may be taking same product from different Agencies... Can you check it out that they have different agency ID. Actually everyone already knows it. That's why there is agency address files so that we can visualize location of agency, clients and routes connecting them. Basically what you can do is take agencies as nodes, and their common Route as edges. Their common client should lie on that route. { Just a hypothesis..}. Also i think there should be negative correlation between demand for common products in such agencies (i.e. two agencies serving on same route , supplying same product to same client)
\n\n[/quote]
\n\nI'm not sure that's the case, if so I feel it should have been mentioned. In addition how would we then go about predicting with any type of accuracy which agency they will order from next week. Another post on the forum makes me think that agency may mean a store location.
\n\n\n\nBut the description mentions it as a sales depot. I would take that as the source of product not the destination.
\n\nI know my problem stems from the use of only 3 of the 6 columns as unique identifiers(keys) ; Semana, Agencia_ID, Canal_ID, Ruta_SAK, Cliente_ID, Producto_ID. Really the problem is we are expected to have some type of operational understand to how Group Bimbo works, which we don't and isn't easy to find.
\n\nIf they could just explain a little about what these fields represent and how they are related it would allow a much greater understanding of the actual problem.
", "rawMarkdown": "[quote=Shahnawaz Akhtar;124476]\r\n\r\nSame client may be taking same product from different Agencies... Can you check it out that they have different agency ID. Actually everyone already knows it. That's why there is agency address files so that we can visualize location of agency, clients and routes connecting them. Basically what you can do is take agencies as nodes, and their common Route as edges. Their common client should lie on that route. { Just a hypothesis..}. Also i think there should be negative correlation between demand for common products in such agencies (i.e. two agencies serving on same route , supplying same product to same client) \r\n\r\n[/quote]\r\n\r\nI'm not sure that's the case, if so I feel it should have been mentioned. In addition how would we then go about predicting with any type of accuracy which agency they will order from next week. Another post on the forum makes me think that agency may mean a store location.\r\n\r\nhttps://www.kaggle.com/c/grupo-bimbo-inventory-demand/forums/t/21601/what-are-we-actually-predicting-here/123470#post123470 \r\n\r\nBut the description mentions it as a sales depot. I would take that as the source of product not the destination. \r\n\r\nI know my problem stems from the use of only 3 of the 6 columns as unique identifiers(keys) ; Semana, Agencia_ID, Canal_ID, Ruta_SAK, Cliente_ID, Producto_ID. Really the problem is we are expected to have some type of operational understand to how Group Bimbo works, which we don't and isn't easy to find. \r\n\r\nIf they could just explain a little about what these fields represent and how they are related it would allow a much greater understanding of the actual problem.", "votes": null }, { "id": "124686", "postDate": "06/21/2016 09:28:35", "content": "There are no duplicates if you use all columns.
", "rawMarkdown": "There are no duplicates if you use all columns.", "votes": null }, { "id": "124710", "postDate": "06/21/2016 14:16:08", "content": "@Kyle7Day, sorry if that comment made you think Agencias (Sales depots) are store locations. Some Agencias sell returned items but most likely those stores have their own Cliente_ID.
\n\nThere's a lot of operational details about Bimbo (both in English and Spanish). Every year they release a report to their investors, and are indexed in a few search engines. There are other articles and reports from third parties (academia) about their distribution strategies.
", "rawMarkdown": "Kyle7Day, sorry if that comment made you think Agencias (Sales depots) are store locations. Some Agencias sell returned items but most likely those stores have their own Cliente_ID. \r\n\r\nThere's a lot of operational details about Bimbo (both in English and Spanish). Every year they release a report to their investors, and are indexed in a few search engines. There are other articles and reports from third parties (academia) about their distribution strategies.", "votes": null }, { "id": "127667", "postDate": "07/14/2016 20:53:15", "content": "Please ignore this, I found an error in my sql that created my result_table.
\n\nI was looking for tuples with data for 7 weeks, and found that there are tuples with additional weeks. The following query shows that are records with additional weeks:
\n\n SELECT [Agencia_ID],[Ruta_SAK],[Canal_ID],[Cliente_ID],[Producto_ID], Semana, Venta_uni_hoy, Venta_hoy, Dev_uni_proxima, Dev_proxima, Demanda_uni_equil\n FROM [TPE_DS_Target_POC].[dbo].[result_table]\n where [Agencia_ID] = '2242'\n and [Ruta_SAK] = '1003'\n and [Canal_ID] = '1'\n and [Cliente_ID] = '352983'\n and [Producto_ID] = '1109'\n order by semana\n\nAgencia_ID Ruta_SAK Canal_ID Cliente_ID Producto_ID Semana Venta_uni_hoy Venta_hoy Dev_uni_proxima Dev_proxima Demanda_uni_equil\n2242 1003 1 352983 1109 3 18 270.18 0 0.00 18\n2242 1003 1 352983 1109 4 20 300.20 0 0.00 20\n2242 1003 1 352983 1109 5 14 210.14 0 0.00 14\n2242 1003 1 352983 1109 6 6 90.06 4 60.04 2\n2242 1003 1 352983 1109 6 8 120.08 0 0.00 8\n2242 1003 1 352983 1109 7 5 75.05 0 0.00 5\n2242 1003 1 352983 1109 8 19 285.19 0 0.00 19\n2242 1003 1 352983 1109 9 12 180.12 7 105.07 5\nSame client may be taking same product from different Agencies... Can you check it out that they have different agency ID. Actually everyone already knows it. That's why there is agency address files so that we can visualize location of agency, clients and routes connecting them. Basically what you can do is take agencies as nodes, and their common Route as edges. Their common client should lie on that route. { Just a hypothesis..}. Also i think there should be negative correlation between demand for common products in such agencies (i.e. two agencies serving on same route , supplying same product to same client)
", "votes": null, "replies": [] }, { "id": 124477, "author_name": "kyle7day", "author_url": "", "post_date": "06/19/2016 03:57:05", "content": "[quote=Shahnawaz Akhtar;124476]
\n\nSame client may be taking same product from different Agencies... Can you check it out that they have different agency ID. Actually everyone already knows it. That's why there is agency address files so that we can visualize location of agency, clients and routes connecting them. Basically what you can do is take agencies as nodes, and their common Route as edges. Their common client should lie on that route. { Just a hypothesis..}. Also i think there should be negative correlation between demand for common products in such agencies (i.e. two agencies serving on same route , supplying same product to same client)
\n\n[/quote]
\n\nI'm not sure that's the case, if so I feel it should have been mentioned. In addition how would we then go about predicting with any type of accuracy which agency they will order from next week. Another post on the forum makes me think that agency may mean a store location.
\n\n\n\nBut the description mentions it as a sales depot. I would take that as the source of product not the destination.
\n\nI know my problem stems from the use of only 3 of the 6 columns as unique identifiers(keys) ; Semana, Agencia_ID, Canal_ID, Ruta_SAK, Cliente_ID, Producto_ID. Really the problem is we are expected to have some type of operational understand to how Group Bimbo works, which we don't and isn't easy to find.
\n\nIf they could just explain a little about what these fields represent and how they are related it would allow a much greater understanding of the actual problem.
", "votes": null, "replies": [] }, { "id": 124686, "author_name": "cpmpml", "author_url": "", "post_date": "06/21/2016 09:28:35", "content": "There are no duplicates if you use all columns.
", "votes": null, "replies": [] }, { "id": 124710, "author_name": "juanrdz", "author_url": "", "post_date": "06/21/2016 14:16:08", "content": "@Kyle7Day, sorry if that comment made you think Agencias (Sales depots) are store locations. Some Agencias sell returned items but most likely those stores have their own Cliente_ID.
\n\nThere's a lot of operational details about Bimbo (both in English and Spanish). Every year they release a report to their investors, and are indexed in a few search engines. There are other articles and reports from third parties (academia) about their distribution strategies.
", "votes": null, "replies": [] }, { "id": 127667, "author_name": "jorgepalma", "author_url": "", "post_date": "07/14/2016 20:53:15", "content": "Please ignore this, I found an error in my sql that created my result_table.
\n\nI was looking for tuples with data for 7 weeks, and found that there are tuples with additional weeks. The following query shows that are records with additional weeks:
\n\n SELECT [Agencia_ID],[Ruta_SAK],[Canal_ID],[Cliente_ID],[Producto_ID], Semana, Venta_uni_hoy, Venta_hoy, Dev_uni_proxima, Dev_proxima, Demanda_uni_equil\n FROM [TPE_DS_Target_POC].[dbo].[result_table]\n where [Agencia_ID] = '2242'\n and [Ruta_SAK] = '1003'\n and [Canal_ID] = '1'\n and [Cliente_ID] = '352983'\n and [Producto_ID] = '1109'\n order by semana\n\nAgencia_ID Ruta_SAK Canal_ID Cliente_ID Producto_ID Semana Venta_uni_hoy Venta_hoy Dev_uni_proxima Dev_proxima Demanda_uni_equil\n2242 1003 1 352983 1109 3 18 270.18 0 0.00 18\n2242 1003 1 352983 1109 4 20 300.20 0 0.00 20\n2242 1003 1 352983 1109 5 14 210.14 0 0.00 14\n2242 1003 1 352983 1109 6 6 90.06 4 60.04 2\n2242 1003 1 352983 1109 6 8 120.08 0 0.00 8\n2242 1003 1 352983 1109 7 5 75.05 0 0.00 5\n2242 1003 1 352983 1109 8 19 285.19 0 0.00 19\n2242 1003 1 352983 1109 9 12 180.12 7 105.07 5\n