I have a dataset consisting of 5 numerical variables, like the sample in the following image: enter image description here

In this dataset the first four variables are the input and the GDP is the output. I am trying to build a neural network for regression of the GDP variable. For the input variables, I have used PCA to normalize the data, while for the GDP variable I have used a MinMax normalization algorithm. However, the accuracy of the results of the neural network is very low and I guess that the problem is on the normalization of my output variable. Any suggestions on how I can improve the accuracy of my neural network?

  • 1
    $\begingroup$ Why to normalise output variable. Normalisation is perfomed on dependent variable not on indipendent variable. $\endgroup$
    – g_p
    Apr 23 '20 at 20:18

Normally, you shouldn't make normalization on your target data. If you want to increase your accuracy of model, you can try different neural network architectures or activation functions and monitor your train/val/test scores.

If you insist to normalize target data, then you may want to convert your predictions and then calculate the accuracy.

Steps may be like below:

normalize data => fit model => take predictions => denormalize predictions => calculate accuracy

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.