How to predict next value using time series that have only 1 column and 14000 rows ?

2 次查看(过去 30 天)
I want to predict next internet traffic data on links. I have a time series that only 1 column/14000 rows. I want to create a neural network, then I will train it with %70, no validate(if not necessary) and test with %30 of time series values. At last, I want to calculate the MAPE for predictions/target only. I have a few questions, will be glad with your answer:
1- I am not sure but using the NAR network type now. Which network type should I choose ? NARX, NAR or Non Input/Output type ?
2- I think the MAPE is calculating for all 14000 values. But I want to calculate the MAPE for only Predictions/Test(~=4200 values). How can I do that ?
This is my time series. Thank for all answers.

采纳的回答

Greg Heath
Greg Heath 2017-4-7
1. Transpose to get 1 row.
2. Search the NEWSGROUP and ANSWERS with
narnet tutorial
Good Luck,
Greg

更多回答(0 个)

类别

Help CenterFile Exchange 中查找有关 Sequence and Numeric Feature Data Workflows 的更多信息

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by