Powerful and Intuitive: IBM SPSS Decision Trees
Keith McCormick, Jesus Salcedo · 2017
This chapter discusses how IBM SPSS Decision Trees offers four methods, including CHAID, Exhaustive CHAID, CRT, and QUEST. The C5.0 Tree extension command offers a fifth possible option. View could result in a different tree, as the algorithm will treat nominal, ordinal, and continuous independent variables differently. Scale variables pose an interesting problem because Chi-Sq is not designed to investigate scale variables. CHAID buckets scale variables into deciles, and then treats them as ordinal variables. A half dozen or even a dozen different versions of settings for just the CHAID algorithm would not be unusual. CRT has produced a considerably more complex tree. Its performance is about the same as the second CHAID attempt. The original developers of the CRT algorithm approach experimented with a number of variations on the algorithm and discovered that the best results were found when they grew a tree aggressively and pruned it back. The Scoring Wizard is found in the Utilities menu, which lists any .xml files that can be used.