We are predicting the natural log of the sum of all transactions per user. Once the data is updated, as noted above, this will be for all users in test_v2.csv for December 1st, 2018 to January 31st, 2019. For every user in the test set, the target is:\nyuser=∑i=1ntransactionuseri\ntargetuser=ln(yuser+1)\nNote that the dataset does NOT contain data for December 1st 2018 to January 31st 2019. You must identify the unique fullVisitorIds in the provided test_v2.csv and make predictions for them for those unseen months.
\n\nHowever, I see everyone is using all the information from test data set to make the prediction. I believe the result from those predictions are the total amount of transaction revenue from test set time interval, not the prediction for December 1st 2018 to January 31st 2019.
\n\nPlease help me with this. Thank you guys!
", "messages": [ { "id": "411166", "postDate": "10/27/2018 14:32:25", "content": "We are predicting the natural log of the sum of all transactions per user. Once the data is updated, as noted above, this will be for all users in test_v2.csv for December 1st, 2018 to January 31st, 2019. For every user in the test set, the target is:\nyuser=∑i=1ntransactionuseri\ntargetuser=ln(yuser+1)\nNote that the dataset does NOT contain data for December 1st 2018 to January 31st 2019. You must identify the unique fullVisitorIds in the provided test_v2.csv and make predictions for them for those unseen months.
\n\nHowever, I see everyone is using all the information from test data set to make the prediction. I believe the result from those predictions are the total amount of transaction revenue from test set time interval, not the prediction for December 1st 2018 to January 31st 2019.
\n\nPlease help me with this. Thank you guys!
", "rawMarkdown": "# From the Data section:\n# What am I predicting part:\nWe are predicting the natural log of the sum of all transactions per user. Once the data is updated, as noted above, this will be for all users in test_v2.csv for December 1st, 2018 to January 31st, 2019. For every user in the test set, the target is:\nyuser=∑i=1ntransactionuseri\ntargetuser=ln(yuser+1)\nNote that the dataset does NOT contain data for December 1st 2018 to January 31st 2019. You must identify the unique fullVisitorIds in the provided test_v2.csv and make predictions for them for those unseen months.\n\nHowever, I see everyone is using all the information from test data set to make the prediction. I believe the result from those predictions are the total amount of transaction revenue from test set time interval, not the prediction for December 1st 2018 to January 31st 2019.\n\nPlease help me with this. Thank you guys!", "votes": null }, { "id": "411473", "postDate": "10/28/2018 08:41:32", "content": "I have the same confusion. Everyone is predicting revenue for the test data whereas competition requires forecast for unseen months
", "rawMarkdown": "I have the same confusion. Everyone is predicting revenue for the test data whereas competition requires forecast for unseen months", "votes": null }, { "id": "411565", "postDate": "10/28/2018 13:27:03", "content": "I went through the Data section again. The new data will be provide by 11/09, and after new data being uploaded we should do the forward looking prediction. Therefore, I am guessing, with the old data what we have to do is get the total revenue with session information, and with new data we are going to do the prediction without session information.
", "rawMarkdown": "I went through the Data section again. The new data will be provide by 11/09, and after new data being uploaded we should do the forward looking prediction. Therefore, I am guessing, with the old data what we have to do is get the total revenue with session information, and with new data we are going to do the prediction without session information.", "votes": null }, { "id": "412146", "postDate": "10/29/2018 16:53:19", "content": "So how exactly public LB score will be calculated if the data will be available in the future? You will upload your submission and then what, no score? So what's the point to upload more than one submission?
", "rawMarkdown": "So how exactly public LB score will be calculated if the data will be available in the future? You will upload your submission and then what, no score? So what's the point to upload more than one submission?", "votes": null }, { "id": "412782", "postDate": "10/30/2018 19:03:49", "content": "I am confused too.
\n\nAfter Acceptance Deadline we'll wait until January 31st 2019, then they evaluate the natural log of the sum of all transactions per user for real values for December 1st, 2018 to January 31st, 2019, then our submissions will be checked, if it coincides with real values by RMSE score. And then they'll find the leaders and award them.
\n\nAm I right?
", "rawMarkdown": "I am confused too.\n\nAfter Acceptance Deadline we'll **wait until January 31st 2019**, then they evaluate the natural log of the sum of all transactions per user **for real values for December 1st, 2018 to January 31st, 2019**, then our submissions will be checked, if it coincides with real values by RMSE score. And then they'll find the leaders and award them.\n\nAm I right?", "votes": null }, { "id": "412896", "postDate": "10/31/2018 01:24:03", "content": "I think you are right. God it will take so much longer than other competitions to see our result!
", "rawMarkdown": "I think you are right. God it will take so much longer than other competitions to see our result!", "votes": null }, { "id": "412897", "postDate": "10/31/2018 01:24:33", "content": "I think you can refer to Vladimir Smirnov's comment
", "rawMarkdown": "I think you can refer to Vladimir Smirnov's comment", "votes": null }, { "id": "414500", "postDate": "11/02/2018 22:22:37", "content": "Thank you so much for reply and clarification.
", "rawMarkdown": "Thank you so much for reply and clarification.", "votes": null }, { "id": "415672", "postDate": "11/05/2018 13:23:48", "content": "I am still confused,
\n\ntest_v2.csv - the updated test set - contains user transactions from May 1st 2018 to October 15th 2018.
\n\nSo we should suppose all features of the customers (e.g. browser etc) won't change during （01052018,15102018）and （01122018,13012018），then predict the value during（01122018,13012018）。\nThe LB score is not the final score, only reference. Am I right?\nThank you.
", "rawMarkdown": "I am still confused, \n\ntest_v2.csv - the updated test set - contains user transactions from May 1st 2018 to October 15th 2018.\n\nSo we should suppose all features of the customers (e.g. browser etc) won't change during （01052018,15102018）and （01122018,13012018），then predict the value during（01122018,13012018）。\nThe LB score is not the final score, only reference. Am I right?\nThank you.", "votes": null }, { "id": "415840", "postDate": "11/05/2018 19:28:57", "content": "I think we are going to use this test2_v2.csv to predict revenue for each client from December 1st, 2018 to January 31st, 2019
", "rawMarkdown": "I think we are going to use this test2_v2.csv to predict revenue for each client from December 1st, 2018 to January 31st, 2019", "votes": null }, { "id": "416080", "postDate": "11/06/2018 06:46:51", "content": "Yes. \nFirst to predict if they will return during December 1st, 2018 to January 31st, 2019.\nThen predict how much they will spend.\nI guess it's the effective way to deduce MSE.
", "rawMarkdown": "Yes. \nFirst to predict if they will return during December 1st, 2018 to January 31st, 2019.\nThen predict how much they will spend.\nI guess it's the effective way to deduce MSE.", "votes": null } ], "comments": [ { "id": 411473, "author_name": "thaque2050", "author_url": "", "post_date": "10/28/2018 08:41:32", "content": "I have the same confusion. Everyone is predicting revenue for the test data whereas competition requires forecast for unseen months
", "votes": null, "replies": [ { "id": 411565, "author_name": "wzy1993", "author_url": "", "post_date": "10/28/2018 13:27:03", "content": "I went through the Data section again. The new data will be provide by 11/09, and after new data being uploaded we should do the forward looking prediction. Therefore, I am guessing, with the old data what we have to do is get the total revenue with session information, and with new data we are going to do the prediction without session information.
", "votes": null, "replies": [] }, { "id": 414500, "author_name": "thaque2050", "author_url": "", "post_date": "11/02/2018 22:22:37", "content": "Thank you so much for reply and clarification.
", "votes": null, "replies": [] } ] }, { "id": 412146, "author_name": "jacekpoplawski", "author_url": "", "post_date": "10/29/2018 16:53:19", "content": "So how exactly public LB score will be calculated if the data will be available in the future? You will upload your submission and then what, no score? So what's the point to upload more than one submission?
", "votes": null, "replies": [ { "id": 412897, "author_name": "wzy1993", "author_url": "", "post_date": "10/31/2018 01:24:33", "content": "I think you can refer to Vladimir Smirnov's comment
", "votes": null, "replies": [] } ] }, { "id": 412782, "author_name": "etosodetails", "author_url": "", "post_date": "10/30/2018 19:03:49", "content": "I am confused too.
\n\nAfter Acceptance Deadline we'll wait until January 31st 2019, then they evaluate the natural log of the sum of all transactions per user for real values for December 1st, 2018 to January 31st, 2019, then our submissions will be checked, if it coincides with real values by RMSE score. And then they'll find the leaders and award them.
\n\nAm I right?
", "votes": null, "replies": [ { "id": 412896, "author_name": "wzy1993", "author_url": "", "post_date": "10/31/2018 01:24:03", "content": "I think you are right. God it will take so much longer than other competitions to see our result!
", "votes": null, "replies": [] }, { "id": 415672, "author_name": "elulue", "author_url": "", "post_date": "11/05/2018 13:23:48", "content": "I am still confused,
\n\ntest_v2.csv - the updated test set - contains user transactions from May 1st 2018 to October 15th 2018.
\n\nSo we should suppose all features of the customers (e.g. browser etc) won't change during （01052018,15102018）and （01122018,13012018），then predict the value during（01122018,13012018）。\nThe LB score is not the final score, only reference. Am I right?\nThank you.
", "votes": null, "replies": [] }, { "id": 415840, "author_name": "wzy1993", "author_url": "", "post_date": "11/05/2018 19:28:57", "content": "I think we are going to use this test2_v2.csv to predict revenue for each client from December 1st, 2018 to January 31st, 2019
", "votes": null, "replies": [] }, { "id": 416080, "author_name": "elulue", "author_url": "", "post_date": "11/06/2018 06:46:51", "content": "Yes. \nFirst to predict if they will return during December 1st, 2018 to January 31st, 2019.\nThen predict how much they will spend.\nI guess it's the effective way to deduce MSE.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "411166": "# From the Data section:\n# What am I predicting part:\nWe are predicting the natural log of the sum of all transactions per user. Once the data is updated, as noted above, this will be for all users in test_v2.csv for December 1st, 2018 to January 31st, 2019. For every user in the test set, the target is:\nyuser=∑i=1ntransactionuseri\ntargetuser=ln(yuser+1)\nNote that the dataset does NOT contain data for December 1st 2018 to January 31st 2019. You must identify the unique fullVisitorIds in the provided test_v2.csv and make predictions for them for those unseen months.\n\nHowever, I see everyone is using all the information from test data set to make the prediction. I believe the result from those predictions are the total amount of transaction revenue from test set time interval, not the prediction for December 1st 2018 to January 31st 2019.\n\nPlease help me with this. Thank you guys!", "411473": "I have the same confusion. Everyone is predicting revenue for the test data whereas competition requires forecast for unseen months", "411565": "I went through the Data section again. The new data will be provide by 11/09, and after new data being uploaded we should do the forward looking prediction. Therefore, I am guessing, with the old data what we have to do is get the total revenue with session information, and with new data we are going to do the prediction without session information.", "412146": "So how exactly public LB score will be calculated if the data will be available in the future? You will upload your submission and then what, no score? So what's the point to upload more than one submission?", "412782": "I am confused too.\n\nAfter Acceptance Deadline we'll **wait until January 31st 2019**, then they evaluate the natural log of the sum of all transactions per user **for real values for December 1st, 2018 to January 31st, 2019**, then our submissions will be checked, if it coincides with real values by RMSE score. And then they'll find the leaders and award them.\n\nAm I right?", "412896": "I think you are right. God it will take so much longer than other competitions to see our result!", "412897": "I think you can refer to Vladimir Smirnov's comment", "414500": "Thank you so much for reply and clarification.", "415672": "I am still confused, \n\ntest_v2.csv - the updated test set - contains user transactions from May 1st 2018 to October 15th 2018.\n\nSo we should suppose all features of the customers (e.g. browser etc) won't change during （01052018,15102018）and （01122018,13012018），then predict the value during（01122018,13012018）。\nThe LB score is not the final score, only reference. Am I right?\nThank you.", "415840": "I think we are going to use this test2_v2.csv to predict revenue for each client from December 1st, 2018 to January 31st, 2019", "416080": "Yes. \nFirst to predict if they will return during December 1st, 2018 to January 31st, 2019.\nThen predict how much they will spend.\nI guess it's the effective way to deduce MSE." }, "source": "meta" }