I need to ask this question before I go to sleep
Hope when I wake up there will be some answers :)
I have around 80 variables as inputs to my Neural Network. I am wondering if some of these variables highly correlates with others, say with coeff > 0.9, can I just use one and drop the others (as they highly correlate with one another).
Thnak you. Good night
Hope you are awake! ;)
If you are gonna wanna pick one from those highly correlated ones (>0.9), and if the comparison was between the observed or actual and predicted values, I suppose that it is possible.
Dear Kamal,
You should definitely use highly correlated variables as inputs of the network instead of using all of them, coeff > 0.5 often is good for the network.
Check this article for more details:
www.sciencedirect.com/science/article/pii/S1877050918312754/pdf?md5=e3fa8f504daedd44c326e198036b2c66&pid=1-s2.0-S1877050918312754-main.pdf
Hope it helps.
- Badges
- Science topic