Course Overview

Advanced Predictive Modeling using IBM SPSS Modeler (V15). This ILT/ILO course teaches a number of advanced modeling techniques to predict categorical and continuous targets, and is meant for users of IBM SPSS Modeler responsible for building predictive (or “classification “) models. Before reviewing the various modeling techniques, prerequisites for building successful models are addressed. The next lessons discuss Decision List, Support Vector Machines and Bayes Nets. After having discussed individual models, it is demonstrated how multiple models can be combined to improve the predictive power. Finally, the focus is on how to automate the process of finding the best predictive model. Each lesson is accompanied by demonstrations and learning activities, to acquire hands-on experience.

Who Should Attend

This advanced course is for users of IBM SPSS Modeler responsible for building predictive (or “classification”) models

Course Certifications

This course is part of the following Certifications:


You should:

  • Complete Introduction to IBM SPSS Modeler and Data Mining or experience in analyzing data with IBM SPSS Modeler.
  • Be familiar with basic modeling techniques, either through completion of the courses Classifying Customers using IBM SPSS Modeler, and Predicting Continuous Targets using IBM SPSS Modeler, or by experience with IBM SPSS Modeler.

Course Content

Course Introduction

  • Course objectives
  • Course description
  • Course assumptions


Preparing Data for Modeling

  • General data quality issues
  • The Anomaly node
  • The Feature Selection node
  • The Partition node
  • The Balance node


Rule Induction using Decision List

  • Interactive Decision List
  • Direct Decision List


Machine Learning Models

  • Support Vector Machines
  • Bayes Net


Combining Predictive Models

  • The Ensemble node
  • Meta-level modeling
  • Error modeling


Finding the Best Predictive Model

  • The Auto Classifier node
  • The Auto Numeric node


Data Reduction using PCA/Factor

  • Using PCA/Factor to reduce data