![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F801574%2F66e5acc637ca16f322fd46abaf0f7b42%2Fno_0_reading.JPG?generation=1575265033541804&alt=media\")
Hi all,
\n\nThis may have been discussed or I just couldn't find the answer for the meaning behind 0.0 meter reading in the training set. I also wonder in the score of the test set, do we expect 0.0 reading values too?
\n\nFor me they are 2 separate problems if 0.0 is expected in the prediction? Is it because of the failed sensors? or any other reasons?
\n\nThank you.
\n\n[No 0 reading at Building ID 0]\n
Hi all,
\n\nThis may have been discussed or I just couldn't find the answer for the meaning behind 0.0 meter reading in the training set. I also wonder in the score of the test set, do we expect 0.0 reading values too?
\n\nFor me they are 2 separate problems if 0.0 is expected in the prediction? Is it because of the failed sensors? or any other reasons?
\n\nThank you.
\n\n[No 0 reading at Building ID 0]\n
You can find a similar question here.
", "rawMarkdown": "You can find a similar question [here](https://www.kaggle.com/c/ashrae-energy-prediction/discussion/113054).", "votes": null }, { "id": "687153", "postDate": "12/04/2019 01:39:49", "content": "I think the answer is \"it depends\".
\n\nSeems like some 0.0 are legitimate - for example there are a number of swimming pools that pretty much show only summer energy. Some 0.0 there look to be the real deal.
\n\nOthers seem to be very clearly failed sensor readings. The site_0 early readings are for sure not real - that site was not built in 2016.
\n\nWhen you consider that the meters are spread across 1400+ buildings and 16 sites - than there's going to be a nice mix of real and not real. This is a super dirty data set - getting it clean would seem to be a high percentage of the problem.
", "rawMarkdown": "I think the answer is \"it depends\".\n\nSeems like some 0.0 are legitimate - for example there are a number of swimming pools that pretty much show only summer energy. Some 0.0 there look to be the real deal.\n\nOthers seem to be very clearly failed sensor readings. The site_0 early readings are for sure not real - that site was not built in 2016.\n\nWhen you consider that the meters are spread across 1400+ buildings and 16 sites - than there's going to be a nice mix of real and not real. This is a super dirty data set - getting it clean would seem to be a high percentage of the problem.", "votes": null }, { "id": "687246", "postDate": "12/04/2019 05:42:38", "content": "I agree with @pcjimmmy , it depends. If it oscillates between 0 and some small value, it is probably correct. But, if it looks like this, clearly something is wrong. It doesn't look natural.
\n\n![\"1\"](\"https://i.ibb.co/2FYZnjr/12.jpg\")
You can find a similar question here.
", "votes": null, "replies": [] }, { "id": 687153, "author_name": "pcjimmmy", "author_url": "", "post_date": "12/04/2019 01:39:49", "content": "I think the answer is \"it depends\".
\n\nSeems like some 0.0 are legitimate - for example there are a number of swimming pools that pretty much show only summer energy. Some 0.0 there look to be the real deal.
\n\nOthers seem to be very clearly failed sensor readings. The site_0 early readings are for sure not real - that site was not built in 2016.
\n\nWhen you consider that the meters are spread across 1400+ buildings and 16 sites - than there's going to be a nice mix of real and not real. This is a super dirty data set - getting it clean would seem to be a high percentage of the problem.
", "votes": null, "replies": [] }, { "id": 687246, "author_name": "gunesevitan", "author_url": "", "post_date": "12/04/2019 05:42:38", "content": "I agree with @pcjimmmy , it depends. If it oscillates between 0 and some small value, it is probably correct. But, if it looks like this, clearly something is wrong. It doesn't look natural.
\n\n