MATLAB Answers

NARX model gives high accuracy but prediction of other data is too low

6 views (last 30 days)
Murat Yetkin
Murat Yetkin on 22 Sep 2020
Commented: Aditya Patil on 20 Oct 2020
I have created a NARX model using 1 external input and 1 target data set in time series. When I train the system with very long memory lengths for output, I got good accuracy for training but when I try to predict other data set from same system, prediction accuracy is very low.
then I combined 3 datasets sequencelly to train the network and then tried to predict each other seperately. let's say my dataset numbers are 1, 2 and 3, respectively. when I try to predict I got 79% accuracy for 3, 70% accuracy for 2 and 20% acccuracy for 1.
What is the reason of low accuracy for the dataset 1?

  0 Comments

Sign in to comment.

Answers (1)

Aditya Patil
Aditya Patil on 25 Sep 2020
Generally, when you get good results when training, but poor results on test dataset, it means your model is overfitting. There are several techniques to improve accuracy in such situation. To get started, check out the Improve Shallow Neural Network Generalization and Avoid Overfitting documentation.

  2 Comments

Aditya Patil
Aditya Patil on 20 Oct 2020
If you get low accuracy for training data, it means the model is not performing well.

Sign in to comment.

Community Treasure Hunt

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

Start Hunting!