The HPSPLIT Procedure

Example 61.5 Assessing Variable Importance

This example creates a classification tree model to determine important variables (parameters) during the manufacture of a semiconductor device. Some of the variables that are involved in the manufacturing process are as follows: gTemp is the growth temperature of substrate, aTemp is the anneal temperature, Rot is rotation speed, Dopant is the atom that is used during device growth, and Usable indicates whether the device is usable.

The following statements create a data set named MBE_Data, which contains measurements for 20 devices:

data MBE_Data;
   label gTemp  = 'Growth Temperature of Substrate';
   label aTemp  = 'Anneal Temperature';
   label Rot    = 'Rotation Speed';
   label Dopant = 'Dopant Atom';
   label Usable = 'Experiment Could Be Performed';

   input gTemp aTemp Rot Dopant $ 39-40 Usable $ 47-54;
   datalines;
   384.614    633.172    1.01933      C       Unusable
   363.874    512.942    0.72057      C       Unusable
   397.395    671.179    0.90419      C       Unusable
   389.962    653.940    1.01417      C       Unusable
   387.763    612.545    1.00417      C       Unusable
   394.206    617.021    1.07188      Si      Usable
   387.135    616.035    0.94740      Si      Usable
   428.783    745.345    0.99087      Si      Unusable
   399.365    600.932    1.23307      Si      Unusable
   455.502    648.821    1.01703      Si      Unusable
   387.362    697.589    1.01623      Ge      Usable
   408.872    640.406    0.94543      Ge      Usable
   407.734    628.196    1.05137      Ge      Usable
   417.343    612.328    1.03960      Ge      Usable
   482.539    669.392    0.84249      Ge      Unusable
   367.116    564.246    0.99642      Sn      Unusable
   398.594    733.839    1.08744      Sn      Unusable
   378.032    619.561    1.06137      Sn      Usable
   357.544    606.871    0.85205      Sn      Unusable
   384.578    635.858    1.12215      Sn      Unusable
   ;

The following statements create the tree model:

proc hpsplit data=MBE_Data maxdepth=6;
   class Usable Dopant;
   model Usable = gTemp aTemp Rot Dopant;
   prune none;
run;

Output 61.5.1 shows the "Variable Importance" table.

Output 61.5.1: Variable Importance

The HPSPLIT Procedure

Variable Importance
Variable Variable
Label
Training Count
Relative Importance
gTemp Growth Temperature of Substrate 1.0000 2.1022 2
Dopant Dopant Atom 0.7522 1.5811 1
aTemp Anneal Temperature 0.6228 1.3093 1
Rot Rotation Speed 0.3250 0.6831 1



This table shows that the predictor gTemp has the largest value. This means that the growth temperature of substrate is the most important consideration in determining the usability of the sample.