decision tree

[lixuanzhang]

I am trying to build a decision tree, but it takes forever to run, and I have to cancel it since it is stuck with 16% processing. I chose the  target and predictor variables did not change anything in the configuration. I attached a screenshot and the message I got. Any advice would be appreciated. 

[IraWatt]

Hey @lixuanzhang looking at the results window messages you don't have much RAM on your computer. There is this option in the User settings menu to increase Alteryx's available RAM though I am not entirely sure the implications of increasing this:

[lixuanzhang]

I increased to 8200 megabytes a d here is the message I got: 

Info: Designer x64: The Designer x64 reported: Allocating requested memory would be more than available physical memory. Reverting to 1087.8 MB of memory.

How do I allocate more physical memory to the software?

[apathetichell]

Close other ram intensive programs (ie browsers/zoom/teams/Slack etc.). Check out your resource monitor to see what your ram drains are...

[IraWatt]

@lixuanzhang the Alteryx Docs has a section on Memory Usage. The two other settings to look at are the User Settings and Workflow Configuration settings. There is also a similar threat here: Solved: "The machine is running low on available physical ... - Alteryx Community

[lixuanzhang]

My laptop has 8GB memory, and even if I close everything, the decision ree will not run . Now the messages are:

The Designer x64 reported: This is AMP Engine; running 8 worker threads; memory limit 2007.0 MB.

The Designer x64 reported: The following tool ids are using an e1 compatibility layer with 13 conversion point(s): 28, 2, 171, 290, 43, 55, 25, 30, 36, 34, 39, 242, 7

Designer x64 The Designer x64 reported: Running at a Low Priority.

Does that mean I have to get a computer with more memory?

[lixuanzhang]

It finally ran.  The software did not recognize decimals and considered them as strings. After I changed the data type, the model did successfully run.

Also, here are my suggestions for improving the decision tree or all classification techniques.  It would be more valuable if the accuracy, F score etc, etc are reported for the validation dataset. Also, it would be great if a confusion matrix could be automatically generated. Currently,  we have to use formula to get the values for the cells. 

[LiuZhang · Accepted solution]

Decision tree is very fast to run, if you encounter speed or memory issue, then check the data type first as you found in the case.

To answer your later questions, to get a confusion matrix, then model comparison tool does it for you. You can feed in your model and a validation/testing data set, then it will give you a confusion matrix if the target variable is categorical.

[lixuanzhang]

Great!  This works perfectly on the iris dataset. I tried it on another dataset and there are some errors.  Can you tell me what was wrong?  

[LiuZhang]

Without looking at the dataset/workflow, my current guess is data type or spelling mistake somewhere. Make sure there is no extra white spaces.