585 APPLIED ANALYTICS-DECISION TREES

DEPARTMENT OF STATIS TICS
COLLEGE OF ARTS & SCIENCES
330.972.6886 (TELEPHONE)
Course Description
3470:485/585
Applied Analytics-Decision Trees
3 credits
Prerequisite: 3470:461/561 Applied Statistics or Equivalent or permission.
Course Description: Selected topics in predictive modeling using CHAID, classification
and regression trees, logistic regression and neural networks.
Course Topics:
Chi-Squared Independence Test
Set-up and perform a significance test to determine if two categorical variables are
independent or dependent using the Chi-Square distribution.
Data Preparation
Observe and perform data preparation techniques that include data exploration including
data visualization, missing values assessment, and the determination of transformation,
recoding and binning.
Performance of Predictive Models
Assess the performance of a predictive model.
Performance Measures including Risk, Sensitivity, Specificity, and Average Squared Error.
Techniques to compare between different modeling procedures and how to determine the
best model.
CHAID
Predictive modeling CHi-squared Automatic Interaction Detector Trees (CHAID)
technique when the response variable is binary. Build and completely interpret the
CHAID model.
Theoretical background of how the CHAID algorithm works.
Separate the data to be modeled into a Training and Validation datasets.
Pruning techniques to avoid over-fitting of the data.
Perform the CHAID procedure using SPSS.
CART
Perform the Classification and Regression Tree (CART) predictive modeling technique.
Build and completely interpret the CART model.
Understand the theoretical background of how the CART algorithm works.
Determine when it is appropriate to use the CART or CHAID algorithm.
SAS Decision Tree
Perform decision tree modeling techniques using SAS JMP.
Build and completely interpret the decision tree model.
Understand how the decision tree algorithm works in JMP.
Logistic Regression
Learn and understand how to build predictive models using the Logistic Regression
modeling procedure.
Compare the predictive performance of this parametric procedure to other nonparametric decision tree procedures.
Perform the Logistic Regression procedure using SPSS, SAS Enterprise Miner, and SAS
JMP.
SAS Enterprise Miner
Perform both parametric and non-parametric procedures using SAS Enterprise Miner.
Learn the skills necessary to become SAS Enterprise Miner certified. Much of the
material in this certification program overlaps the topics covered in this course.
Additional topics if time allows
Perform the modeling techniques introduced in this course by using the free statistical
software package R.
Perform Artificial Neural Networks modeling procedure using SAS Enterprise Miner.
Perform the parametric modeling procedure Discriminant Analysis as a classification tool.
Fall 2014