Deep Learning doc

tonyboy9

I tried to reproduce the tutorial process Regression using Deep Learning.



I don't understand the Generate Data parameters, so I ran the process with defaults in place. See the Results. I'm trying to understand how to make the Titanic Example Set work in place of Generate Data.

 
Next I tried replacing Generate Data with the sample Titanic. I ran it and got the error message.



My questions are:

Why does the tutorial demonstrate with Generate Data instead of an Example Set?

How can I fix the error message? Which operator do I use to label the Titanic attribute "Survived?"

Thank you for your time.

Tony

jacobcybulski

Note that the Titanic data is use for classification, whereas your initial example was for regression. The Titanic data set has no label, just set Survived as your label, and make sure to change the Performance to Binomial Classification. It will work!

ironfrown

@tonyboy9 you have so many regular attributes because you have converted all nominal attributes to dummy attributes, one per unique value in each attribute. Your label also got split into dummies. The yellow columns are the real result of classification, they indicate the confidence factors for each possible outcome, and are used to determine the prediction value. Have fun. Jacob 

jacobcybulski

Ha ha ha, looks like am talking from two different accounts 😁

jacobcybulski

Note that the Titanic data is use for classification, whereas your initial example was for regression. The Titanic data set has no label, just set Survived as your label, and make sure to change the Performance to Binomial Classification. It will work!

tonyboy9

Thanks again, Jacob. 

To set Survived as label, I went to Set Role.

After that I did some transformation and cleansing.

Finally I changed to Performance Binomial Classification.



I clicked on Run, and to my astonishment the whole thing executed!

Some questions for you about the Results:

At the bottom of the Results, 4 special attributes, why 2,461 regular attributes? There is no Survived attribute.

For the green and yellow columns, I can click on the attributes and get up and down arrows, then back to normal. What does all this give me?

Finally, if you had to guess, how many days would this take writing and debugging code without RapidMiner?

Thanks again.

ironfrown

@tonyboy9 you have so many regular attributes because you have converted all nominal attributes to dummy attributes, one per unique value in each attribute. Your label also got split into dummies. The yellow columns are the real result of classification, they indicate the confidence factors for each possible outcome, and are used to determine the prediction value. Have fun. Jacob 

jacobcybulski

Ha ha ha, looks like am talking from two different accounts 😁