topic Re: Interpreting output from a modelling node in SAS Academy for Data Science

Interpreting output from a modelling node

pvareschi — Mon, 11 May 2020 10:11:01 GMT

Re: Applied Analytics Using SAS Enterprise Miner

I would be grateful if someone could explain the meaning of the following sections in the Output window of a Modelling node (below are screenshots from the Decision Tree based on page 3-84 of the course notes):

1. Classification Table: how are percentages under "Target Percentage" and "Outcome Percentage" calculated?

2. Assessment Score Distribution: are counts under "Number of Events" and "Number of Nonevents" based on the actual values in the Target variable or the predicted classification values/categories?

3. When defining prior probabilities, the values under "Number of Events" and "Number of Nonevents" contain decimal places, as if the observations had been split into fractions (below is a screenshot from the output of the Regression node based on example at page 6.30):

Re: Interpreting output from a modelling node

PaigeMiller — Mon, 11 May 2020 13:32:15 GMT

Target percentage of 59.633 = 1235/(1235+836); Outcome percentage of 50.9909 = 1235/(1235+1187)
Counts are based on predicted probabilities for each observation in the data set
Each record has a predicted probability of being an event (which is not an integer) and a predicted probability of being a non-event (which is not an integer)

Re: Interpreting output from a modelling node

pvareschi — Tue, 12 May 2020 05:06:11 GMT

With regard to point (3), is there any reason why the sum of the "Events", i.e. 1+15.003+127.926+98.220=242.149, does not match the sum of values in variable P_TARGETB1, i.e. 264.2143, across all cases in the Training dataset?