Congrats to winners especially for weiwei. And have a surprise that Faron didn't submit finally.\nThanks for sh1ng's baseline and Faron's F1 optimizer script.\nsjv's DL solution is wonderful,thanks for sharing.
\n\nMODEL:
\n\n1.user product pair model.
\n\n2.if none model.
\n\nFEATURES:
\n\nUser X Order features:
\n\nProducts features:
\n\nAisle features:
\n\nDepartment features:
\n\nUser features:
\n\nUser X Products features:
\n\nUser X Aisles features:
\n\nUser X Department features:
\n\nProducts X time features:
\n\norder_number features:
\n\nDATA:
\n\nI add the last prior as train data.Also,I add the second to the last prior,but not didn't improve much.
\n\nENSEMBLE:
\n\nall[\"label\"]=0.28*all[\"prediction_arbor\"]+0.12*all[\"prediction_lgbm\"]+0.36*all[\"label_2\"]+0.24*all[\"label_1\"]
\n\nprediction_arbor AND prediction_lgbm is from sh1ng's baseline.
\n\nlabel_2 trained with shuffle, label_1 traind split by userid.They have the same features.
\n\nSUB:
\n\nUse faron's script with my \"none model\".
\n\ncode link：https://github.com/plantsgo/Instacart-Market-Basket-Analysis/tree/master
\n\nThanks~and hope for more sharing~
", "comments": [ { "id": 213676, "author_name": "plantsgo", "author_url": "", "post_date": "2017-08-15T03:45:12.937000", "content": "Base features LB:0.4055.\nAdd data LB:0.4064.\nEnsemble LB:0.4075.\nThe rest improvement is overfitting...
", "votes": 1, "replies": [ { "id": 213678, "author_name": "", "author_url": "", "post_date": "2017-08-15T03:48:33.043000", "content": "congratulations..\nHow did you choose the model weights ? and why did you have 2 different models with same features but different splits?
", "votes": 0, "replies": [] }, { "id": 213679, "author_name": "plantsgo", "author_url": "", "post_date": "2017-08-15T03:53:13.150000", "content": "The weights is from LB score.\nActually split random is overfitting,but I found the LB score is better...
", "votes": 0, "replies": [] }, { "id": 215496, "author_name": "Jing Lu", "author_url": "", "post_date": "2017-08-21T22:25:45.210000", "content": "How do you get two decimal number of parameters with leader board score? Do you use some kinds of gradient decent?
", "votes": 0, "replies": [] } ] }, { "id": 216010, "author_name": "百特曼", "author_url": "", "post_date": "2017-08-24T00:18:27.190000", "content": "Did anyone meet this problem:
\n\nValueError: DataFrame.dtypes for data must be int, float or bool. Did not expect the data types in fields aisle_id, days_since_prior_order_list_30_count, department_id, dow_equal, first_order_number, first_order_number_aisle, first_order_number_department, last_order_number, last_order_number_aisle, last_order_number_department, nb_orders, order_dow, order_hour_of_day, order_id, orders_user_aisle, orders_user_department, product_id, reorder_products_count, since_first_order, since_last_order, total_distinct_items, total_items, up_skip_products_max, up_skip_products_min, up_time_gap_mode, user_aisle_product_nunique, user_department_product_nunique, user_product_add_to_cart_order_mode, user_product_count, user_product_dow_mode
", "votes": 0, "replies": [] }, { "id": 215599, "author_name": "YaGana Sheriff-Hussaini", "author_url": "", "post_date": "2017-08-22T09:07:40.820000", "content": "Congratulations @Plantsgo.
", "votes": 0, "replies": [] }, { "id": 215497, "author_name": "Jing Lu", "author_url": "", "post_date": "2017-08-21T22:28:14.350000", "content": "Congrats! How do you think about label_1 and label_2 model? Are they very different from prediction_lgbm?
", "votes": 0, "replies": [] }, { "id": 215149, "author_name": "FuLin", "author_url": "", "post_date": "2017-08-20T04:48:09.017000", "content": "666
\n\nQuick question: are you using the predicted probability for each user-product pair or the log-odds of the pair to combine multiple models?
", "votes": 0, "replies": [ { "id": 215436, "author_name": "plantsgo", "author_url": "", "post_date": "2017-08-21T16:53:42.317000", "content": "yes,I used it as features of the \"none model\"
", "votes": 0, "replies": [] } ] }, { "id": 214147, "author_name": "KALE", "author_url": "", "post_date": "2017-08-16T04:05:29.917000", "content": "Congrats!
", "votes": 0, "replies": [] }, { "id": 213853, "author_name": "piupiu", "author_url": "", "post_date": "2017-08-15T11:50:29.563000", "content": "到此一游~
", "votes": 0, "replies": [] }, { "id": 213849, "author_name": "Jiashen Liu", "author_url": "", "post_date": "2017-08-15T11:38:20.683000", "content": "Congratulations!
", "votes": 0, "replies": [] }, { "id": 213840, "author_name": "🐳鲲(China)", "author_url": "", "post_date": "2017-08-15T11:29:32.297000", "content": "Congrats to you!
", "votes": 0, "replies": [] }, { "id": 213716, "author_name": "viredery", "author_url": "", "post_date": "2017-08-15T05:25:04.470000", "content": "Congrats to you!
", "votes": 0, "replies": [] }, { "id": 213691, "author_name": "百特曼", "author_url": "", "post_date": "2017-08-15T04:19:39.630000", "content": "Thanks for sharing! If I follow your github, can I get the same score in the private leaderboard? And I just walked roughly through your code, I want to figure out if I need to use something like GPU to run your code? Thanks again
", "votes": 0, "replies": [ { "id": 213723, "author_name": "plantsgo", "author_url": "", "post_date": "2017-08-15T05:44:06.467000", "content": "Word2Vec need GPU.Maybe the nearly score,because the \"none model\" I didn't update for it has some features based on \"user_products_model\".Then you need to change some path by yourself.
", "votes": 0, "replies": [] } ] }, { "id": 213674, "author_name": "StuartLu", "author_url": "", "post_date": "2017-08-15T03:42:30.013000", "content": "congrats
", "votes": 0, "replies": [] }, { "id": 213670, "author_name": "豆花儿", "author_url": "", "post_date": "2017-08-15T03:31:54.887000", "content": "沙发没了\nsofa none
", "votes": 0, "replies": [] }, { "id": 213668, "author_name": "hot-boy", "author_url": "", "post_date": "2017-08-15T03:31:04.673000", "content": "", "votes": 0, "replies": [] }, { "id": 213667, "author_name": "hot-boy", "author_url": "", "post_date": "2017-08-15T03:30:58.357000", "content": "Congrats to you!
", "votes": 0, "replies": [] }, { "id": 213665, "author_name": "qianqian", "author_url": "", "post_date": "2017-08-15T03:30:10.987000", "content": "Congrats to you!
", "votes": 0, "replies": [] }, { "id": 213749, "author_name": "CPMP", "author_url": "", "post_date": "2017-08-15T07:07:13.177000", "content": "congrats, and thanks for sharing!
", "votes": 0, "replies": [] }, { "id": 213671, "author_name": "", "author_url": "", "post_date": "2017-08-15T03:37:25.313000", "content": "Thank you so much.
", "votes": 0, "replies": [] } ], "comment_count": 23, "messages": [ { "id": 213663, "postDate": "2017-08-15T03:27:43.107Z", "content": "Congrats to winners especially for weiwei. And have a surprise that Faron didn't submit finally.\nThanks for sh1ng's baseline and Faron's F1 optimizer script.\nsjv's DL solution is wonderful,thanks for sharing.
\n\nMODEL:
\n\n1.user product pair model.
\n\n2.if none model.
\n\nFEATURES:
\n\nUser X Order features:
\n\nProducts features:
\n\nAisle features:
\n\nDepartment features:
\n\nUser features:
\n\nUser X Products features:
\n\nUser X Aisles features:
\n\nUser X Department features:
\n\nProducts X time features:
\n\norder_number features:
\n\nDATA:
\n\nI add the last prior as train data.Also,I add the second to the last prior,but not didn't improve much.
\n\nENSEMBLE:
\n\nall[\"label\"]=0.28*all[\"prediction_arbor\"]+0.12*all[\"prediction_lgbm\"]+0.36*all[\"label_2\"]+0.24*all[\"label_1\"]
\n\nprediction_arbor AND prediction_lgbm is from sh1ng's baseline.
\n\nlabel_2 trained with shuffle, label_1 traind split by userid.They have the same features.
\n\nSUB:
\n\nUse faron's script with my \"none model\".
\n\ncode link：https://github.com/plantsgo/Instacart-Market-Basket-Analysis/tree/master
\n\nThanks~and hope for more sharing~
", "rawMarkdown": "Congrats to winners especially for weiwei. And have a surprise that Faron didn't submit finally.\nThanks for sh1ng's baseline and Faron's F1 optimizer script.\nsjv's DL solution is wonderful,thanks for sharing.\n\n**MODEL:**\n\n1.user product pair model.\n\n2.if none model.\n\n**FEATURES:**\n\nUser X Order features:\n\nProducts features:\n\nAisle features:\n\nDepartment features:\n\nUser features:\n\nUser X Products features:\n\nUser X Aisles features:\n\nUser X Department features:\n\nProducts X time features:\n\norder_number features:\n\n**DATA:**\n\nI add the last prior as train data.Also,I add the second to the last prior,but not didn't improve much.\n\n**ENSEMBLE:**\n\nall[\"label\"]=0.28*all[\"prediction_arbor\"]+0.12*all[\"prediction_lgbm\"]+0.36*all[\"label_2\"]+0.24*all[\"label_1\"]\n\nprediction_arbor AND prediction_lgbm is from sh1ng's baseline.\n\nlabel_2 trained with shuffle, label_1 traind split by userid.They have the same features.\n\n**SUB:**\n\nUse faron's script with my \"none model\".\n\ncode link：https://github.com/plantsgo/Instacart-Market-Basket-Analysis/tree/master\n\nThanks~and hope for more sharing~", "votes": 45 }, { "id": 213676, "postDate": "2017-08-15T03:45:12.937Z", "content": "Base features LB:0.4055.\nAdd data LB:0.4064.\nEnsemble LB:0.4075.\nThe rest improvement is overfitting...
", "rawMarkdown": "Base features LB:0.4055.\nAdd data LB:0.4064.\nEnsemble LB:0.4075.\nThe rest improvement is overfitting...", "votes": 1, "replies": [ { "id": 213678, "postDate": "2017-08-15T03:48:33.043Z", "content": "congratulations..\nHow did you choose the model weights ? and why did you have 2 different models with same features but different splits?
", "rawMarkdown": "congratulations..\nHow did you choose the model weights ? and why did you have 2 different models with same features but different splits?" }, { "id": 213679, "postDate": "2017-08-15T03:53:13.150Z", "content": "The weights is from LB score.\nActually split random is overfitting,but I found the LB score is better...
", "rawMarkdown": "The weights is from LB score.\nActually split random is overfitting,but I found the LB score is better..." }, { "id": 215496, "postDate": "2017-08-21T22:25:45.210Z", "content": "How do you get two decimal number of parameters with leader board score? Do you use some kinds of gradient decent?
", "rawMarkdown": "How do you get two decimal number of parameters with leader board score? Do you use some kinds of gradient decent?" } ] }, { "id": 216010, "postDate": "2017-08-24T00:18:27.190Z", "content": "Did anyone meet this problem:
\n\nValueError: DataFrame.dtypes for data must be int, float or bool. Did not expect the data types in fields aisle_id, days_since_prior_order_list_30_count, department_id, dow_equal, first_order_number, first_order_number_aisle, first_order_number_department, last_order_number, last_order_number_aisle, last_order_number_department, nb_orders, order_dow, order_hour_of_day, order_id, orders_user_aisle, orders_user_department, product_id, reorder_products_count, since_first_order, since_last_order, total_distinct_items, total_items, up_skip_products_max, up_skip_products_min, up_time_gap_mode, user_aisle_product_nunique, user_department_product_nunique, user_product_add_to_cart_order_mode, user_product_count, user_product_dow_mode
", "rawMarkdown": "Did anyone meet this problem:\n\nValueError: DataFrame.dtypes for data must be int, float or bool. Did not expect the data types in fields aisle_id, days_since_prior_order_list_30_count, department_id, dow_equal, first_order_number, first_order_number_aisle, first_order_number_department, last_order_number, last_order_number_aisle, last_order_number_department, nb_orders, order_dow, order_hour_of_day, order_id, orders_user_aisle, orders_user_department, product_id, reorder_products_count, since_first_order, since_last_order, total_distinct_items, total_items, up_skip_products_max, up_skip_products_min, up_time_gap_mode, user_aisle_product_nunique, user_department_product_nunique, user_product_add_to_cart_order_mode, user_product_count, user_product_dow_mode" }, { "id": 215599, "postDate": "2017-08-22T09:07:40.820Z", "content": "Congratulations @Plantsgo.
", "rawMarkdown": "Congratulations @Plantsgo." }, { "id": 215497, "postDate": "2017-08-21T22:28:14.350Z", "content": "Congrats! How do you think about label_1 and label_2 model? Are they very different from prediction_lgbm?
", "rawMarkdown": "Congrats! How do you think about label_1 and label_2 model? Are they very different from prediction_lgbm?" }, { "id": 215149, "postDate": "2017-08-20T04:48:09.017Z", "content": "666
\n\nQuick question: are you using the predicted probability for each user-product pair or the log-odds of the pair to combine multiple models?
", "rawMarkdown": "666\n\nQuick question: are you using the predicted probability for each user-product pair or the log-odds of the pair to combine multiple models? ", "replies": [ { "id": 215436, "postDate": "2017-08-21T16:53:42.317Z", "content": "yes,I used it as features of the \"none model\"
", "rawMarkdown": "yes,I used it as features of the \"none model\"" } ] }, { "id": 214147, "postDate": "2017-08-16T04:05:29.917Z", "content": "Congrats!
", "rawMarkdown": "Congrats!" }, { "id": 213853, "postDate": "2017-08-15T11:50:29.563Z", "content": "到此一游~
", "rawMarkdown": "到此一游~" }, { "id": 213849, "postDate": "2017-08-15T11:38:20.683Z", "content": "Congratulations!
", "rawMarkdown": "Congratulations!" }, { "id": 213840, "postDate": "2017-08-15T11:29:32.297Z", "content": "Congrats to you!
", "rawMarkdown": "Congrats to you!" }, { "id": 213716, "postDate": "2017-08-15T05:25:04.470Z", "content": "Congrats to you!
", "rawMarkdown": "Congrats to you!" }, { "id": 213691, "postDate": "2017-08-15T04:19:39.630Z", "content": "Thanks for sharing! If I follow your github, can I get the same score in the private leaderboard? And I just walked roughly through your code, I want to figure out if I need to use something like GPU to run your code? Thanks again
", "rawMarkdown": "Thanks for sharing! If I follow your github, can I get the same score in the private leaderboard? And I just walked roughly through your code, I want to figure out if I need to use something like GPU to run your code? Thanks again", "replies": [ { "id": 213723, "postDate": "2017-08-15T05:44:06.467Z", "content": "Word2Vec need GPU.Maybe the nearly score,because the \"none model\" I didn't update for it has some features based on \"user_products_model\".Then you need to change some path by yourself.
", "rawMarkdown": "Word2Vec need GPU.Maybe the nearly score,because the \"none model\" I didn't update for it has some features based on \"user_products_model\".Then you need to change some path by yourself." } ] }, { "id": 213674, "postDate": "2017-08-15T03:42:30.013Z", "content": "congrats
", "rawMarkdown": "congrats\n" }, { "id": 213670, "postDate": "2017-08-15T03:31:54.887Z", "content": "沙发没了\nsofa none
", "rawMarkdown": "沙发没了\nsofa none" }, { "id": 213668, "postDate": "2017-08-15T03:31:04.673Z", "rawMarkdown": "" }, { "id": 213667, "postDate": "2017-08-15T03:30:58.357Z", "content": "Congrats to you!
", "rawMarkdown": "Congrats to you!" }, { "id": 213665, "postDate": "2017-08-15T03:30:10.987Z", "content": "Congrats to you!
", "rawMarkdown": "Congrats to you!" }, { "id": 213749, "postDate": "2017-08-15T07:07:13.177Z", "content": "congrats, and thanks for sharing!
", "rawMarkdown": "congrats, and thanks for sharing!" }, { "id": 213671, "postDate": "2017-08-15T03:37:25.313Z", "content": "Thank you so much.
", "rawMarkdown": "Thank you so much." } ], "complete": true, "source": "live", "fetched_at": "2026-10-20T02:02:59Z", "competition_id": 6644, "competition_slug": "basket-analysis", "team_id": 674056, "solution_url": "/competitions/basket-analysis/writeups/12th-solution" }