Hi, I'm new Kaggle and machine learning but how can I submit a prediction for one time period and have it used for another? I don't use kernels, as I like using my own PC, so they won't have access to my code to run it for the new test set right? So I've learned a lot from trying out this competition but I'd love to understand better how we will predict the \" true\" test set.
", "messages": [ { "id": "421230", "postDate": "11/14/2018 18:58:27", "content": "Hi, I'm new Kaggle and machine learning but how can I submit a prediction for one time period and have it used for another? I don't use kernels, as I like using my own PC, so they won't have access to my code to run it for the new test set right? So I've learned a lot from trying out this competition but I'd love to understand better how we will predict the \" true\" test set.
", "rawMarkdown": "Hi, I'm new Kaggle and machine learning but how can I submit a prediction for one time period and have it used for another? I don't use kernels, as I like using my own PC, so they won't have access to my code to run it for the new test set right? So I've learned a lot from trying out this competition but I'd love to understand better how we will predict the \" true\" test set.", "votes": null }, { "id": "421337", "postDate": "11/14/2018 23:16:50", "content": "Hi Frederico,\nYou (same as everyone else, even those who use kernels), will have to predict the \"true\" test set (i.e. revenues for Dec-Jan) WITHOUT using the explanatory variables from Dec-Jan. You have to limit yourself only to train+test datasets that are already available (up to mid October) in the \"Data\" section of this competition. There will be no re-runs of predictive algorithms with future explanatory data.
", "rawMarkdown": "Hi Frederico,\nYou (same as everyone else, even those who use kernels), will have to predict the \"true\" test set (i.e. revenues for Dec-Jan) WITHOUT using the explanatory variables from Dec-Jan. You have to limit yourself only to train+test datasets that are already available (up to mid October) in the \"Data\" section of this competition. There will be no re-runs of predictive algorithms with future explanatory data.", "votes": null }, { "id": "422043", "postDate": "11/15/2018 18:00:34", "content": "So it's a completely different problem then... that we had 15 days to do? So when I submit my table, it's supposed to be predicting the sum of revenue per user during December-January!
", "rawMarkdown": "So it's a completely different problem then... that we had 15 days to do? So when I submit my table, it's supposed to be predicting the sum of revenue per user during December-January!", "votes": null }, { "id": "425297", "postDate": "11/21/2018 12:13:07", "content": "Yes, the problem is quite different now.. On the positive side, the final result will be evaluated on true-out-of-sample dataset, which is something new for kaggle (I haven't seen it before, at least)
", "rawMarkdown": "Yes, the problem is quite different now.. On the positive side, the final result will be evaluated on true-out-of-sample dataset, which is something new for kaggle (I haven't seen it before, at least)", "votes": null } ], "comments": [ { "id": 421337, "author_name": "denysbondar", "author_url": "", "post_date": "11/14/2018 23:16:50", "content": "Hi Frederico,\nYou (same as everyone else, even those who use kernels), will have to predict the \"true\" test set (i.e. revenues for Dec-Jan) WITHOUT using the explanatory variables from Dec-Jan. You have to limit yourself only to train+test datasets that are already available (up to mid October) in the \"Data\" section of this competition. There will be no re-runs of predictive algorithms with future explanatory data.
", "votes": null, "replies": [ { "id": 422043, "author_name": "fredericosantos", "author_url": "", "post_date": "11/15/2018 18:00:34", "content": "So it's a completely different problem then... that we had 15 days to do? So when I submit my table, it's supposed to be predicting the sum of revenue per user during December-January!
", "votes": null, "replies": [] }, { "id": 425297, "author_name": "denysbondar", "author_url": "", "post_date": "11/21/2018 12:13:07", "content": "Yes, the problem is quite different now.. On the positive side, the final result will be evaluated on true-out-of-sample dataset, which is something new for kaggle (I haven't seen it before, at least)
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "421230": "Hi, I'm new Kaggle and machine learning but how can I submit a prediction for one time period and have it used for another? I don't use kernels, as I like using my own PC, so they won't have access to my code to run it for the new test set right? So I've learned a lot from trying out this competition but I'd love to understand better how we will predict the \" true\" test set.", "421337": "Hi Frederico,\nYou (same as everyone else, even those who use kernels), will have to predict the \"true\" test set (i.e. revenues for Dec-Jan) WITHOUT using the explanatory variables from Dec-Jan. You have to limit yourself only to train+test datasets that are already available (up to mid October) in the \"Data\" section of this competition. There will be no re-runs of predictive algorithms with future explanatory data.", "422043": "So it's a completely different problem then... that we had 15 days to do? So when I submit my table, it's supposed to be predicting the sum of revenue per user during December-January!", "425297": "Yes, the problem is quite different now.. On the positive side, the final result will be evaluated on true-out-of-sample dataset, which is something new for kaggle (I haven't seen it before, at least)" }, "source": "meta" }