I want to construct a Neural network regression model for data (8 inputs and 1 output). My problem is that my output has exponential distribution. When I normalized it in 'log scale' or 'min-max scale' error is decreasing but the error in my original-output is rising by 10 times. How to fit this data. Are there any other transformations I can try on my data or any other algorithms or models which will help me.

This is the correlation plot without scaling the output, it has exponential distribution. The last one is the output variable in the plot:

This is the correlation plot for log scaled output which has gaussian distribution, the error is decreasing in this case but still high when converted to original data:



Your Answer

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

Browse other questions tagged or ask your own question.