topic Understand neural network output in EMiner in SAS Data Science

Understand neural network output in EMiner

devmaaza — Tue, 18 Oct 2016 01:51:31 GMT

I have a csv file with 100 rows and 16 variables. My target variable is numerical. I am trying to use neural network to build a model. When I run the model (neural network node), I get some result. However I am unable to understand what that result actually means and how can i use it to predict values. How can I get a table of all the actual target values and the predicted target values?

The model i used is as follow

As I am new to AI and EMiner, any help given will be very appreciated.

Thanks.

Re: Understand neural network output in EMiner

Haikuo — Tue, 18 Oct 2016 13:57:58 GMT

I just started to involve some predictive modeling, this is what I know about Neural network, experts please do chip in.

NN model is known to be difficult on explanations. They are useful but in many cases, we don't know why. One of the popular implementations of neural network is fraud detection in banking , you will get Yes or No for each event, however, you may not be able to explain it to your boss.

If you are aiming at obtaining a traditional model with parameters that can be explained, after first doing a neural network model, you can then try to simulate it using decision tree (regular regresion model seems to be harder to simulate NN model).

Re: Understand neural network output in EMiner

WendyCzika — Tue, 18 Oct 2016 14:38:48 GMT

You can view the predicted target values by clicking on Exported Data from the properties panel (shown below), then "Browse...". The predicted values will be in a column named P_target where target is the name of your target variable. It also might be helpful to view the score code from the Results of the Neural Network node (under View>Scoring>SAS Code) to see the calculations for the hidden units and output units that are used to calculate P_target.

To help you understand what inputs are important in your neural network, see this post:

https://communities.sas.com/t5/SAS-Data-Mining/Interpreting-Neural-Network/m-p/250372/highlight/true#M3705