How many historical records can I use? #25
Replies: 2 comments 2 replies
-
|
It is your own decision. You can take the baseline settings as an example.
All Best,
Xinjiang Lu
…On Sun, Apr 17, 2022 at 6:42 PM springer-exe ***@***.***> wrote:
The report say
"For example, given at 6:00 A.M. today, it is required to eff ectively
forecast the wind power generation beginning from 6:00 A.M. on this day to
5:50 AM on the day after tomorrow, *given a series of historical records*
of the wind farm and the related wind turbines."
How many historical records can I use?
1 day? 2 days?
—
Reply to this email directly, view it on GitHub
<#25>, or
unsubscribe
<https://github.com/notifications/unsubscribe-auth/AEVHY33K4OXH5MIZGWWUQO3VFPTIZANCNFSM5TTWG64A>
.
You are receiving this because you are subscribed to this thread.Message
ID: ***@***.***>
|
Beta Was this translation helpful? Give feedback.
-
|
Hi, there,
The currently released data is dedicated to training and validating your
own model. Of course, you can use it to evaluate your model in the
current stage.
The next-released data for testing begins right after the currently
released data.
The procedure in the upcoming tests probably goes like this:
The setting of the prediction is length-Lx-Length-288, where Lx is the
input length for predicting the next 288 time points.
If the time span of the released test set is 60 days, then the last 30 days
will be evaluated for the overall performance. For instance, if one adopts
a rolling window with stride 1 on the test data, as well as the
length-288-length-288 prediction strategy, then, there will be around
60*24*6 - 288*2 + 1 predictions, and the last 30*24*6 predictions will be
examined to calculate the final score.
The evaluation script will be released in the next phase.
All The Best,
Xinjiang Lu
…On Mon, Apr 18, 2022 at 10:36 AM springer-exe ***@***.***> wrote:
我还是没有太理解啊，最终测试的时候，也是我可以决定的吗？没有一个规范的输入输出样例吗？我提交代码的时候，也是如 baseline
一样，给我一个一模一样的“sdwpf_baidukddcup2022_full.CSV”文件，然后我自己给自己打分吗？
I still don't quite understand. I can decide it in finally test? Don't you
have a standard input and output sample?
—
Reply to this email directly, view it on GitHub
<#25 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AEVHY37V36ZK7TOBZC63B33VFTDDRANCNFSM5TTWG64A>
.
You are receiving this because you commented.Message ID:
***@***.***
com>
|
Beta Was this translation helpful? Give feedback.
-
|
To be honest, I think the length of Lx should be fixed in advance, because for some models, this parameter is very important, such as TCN, we need to design the corresponding number of layers and void rate according to the input length.I hope the organizers will give this request a good thought. |
Beta Was this translation helpful? Give feedback.
-
|
We will take your concern into account. Thanks. |
Beta Was this translation helpful? Give feedback.
-
The report say
"For example, given at 6:00 A.M. today, it is required to eff ectively forecast the wind power generation beginning from 6:00 A.M. on this day to 5:50 AM on the day after tomorrow, given a series of historical records of the wind farm and the related wind turbines."
How many historical records can I use?
1 day? 2 days?
Beta Was this translation helpful? Give feedback.
All reactions