SAS Enterprise Miner Software Course

This course teaches several predictive modeling techniques available in SAS/Enterprise Miner Software.  Participants learn the Enterprise Miner Interface.  Predictive models based on Logistic Regression, Neural Networks and Decision Trees are developed.

This course is designed for anyone who needs to work with SAS/Enterprise Miner software.

Upon completion of this course, the participant will be able to:
> Navigate the Enterprise Miner Interface
> Construct a Process Flow Diagram
> View Distribution Characteristics of Variables
> Transform Input Variables
> Incorporate Sub-Diagrams into Process Flow Diagram
> Create HTML Summary Report of Enterprise Miner Project
> Generate and save SAS code created by Enterprise Miner
> Develop Logistic Regression Model
> Develop Neural Network Multilayer Perceptron Model (MLP)
> Develop Neural Network Cascade Model
> Develop Decision Tree Model
> Use Lift Charts and ROC curves to assess individual models and compare different models
> Score data
> Create composite model from several models

Concepts and terminology
Overall Enterprise Miner capabilities
Interface Components

Create Process Flow Diagram
Adding Nodes
Opening Nodes
Modifying node parameters
Running a diagram

Data Preparation
Select data set
Assign variable roles
Partition data set into different model roles
View variable distribution characteristics
Transform variables
Filter outliers
Imputation techniques to replace missing values

Create Sub diagram
Select nodes to include in sub-diagram
Enter and Exit nodes for sub-diagram
Collapse sub-diagram

Logistic Regression
Default settings
Selection criteria
Setting probabilities for variables to enter and remain in model
Significant variables
Parameter estimates
Misclassification rate and overall model assessment

Neural Networks
Basic concepts and terminology
Construct a Multilayer Perceptron Model
Construct a Neural Network based on Principal Components
Construct a Cascade Neural Network
Modifying underlying architecture
Changing model parameters
Variable weights
Training history
Misclassification rate and overall model assessment

Decision Trees
Basic concepts and terminology
Tree Structures
Splitting Criteria
Pruning
Misclassification rate and overall model assessment

Score Data
Score an independent data set
Save score code to external file
Use score code outside Enterprise Miner

Assess Model
Assess individual models
Compare models
Lift chart
ROC curve

Composite Model
Combine several models to form composite model

SAS programming experience and the understanding of basic statistical concepts and linear and logistic regression are required.