Predictive Modelling using Logistic Regression

Duration: 2 days

If you experience any problems with our on-line booking service, you can call us on +44 (0)131 225 6952.

Description

This course is designed for predictive modellers and data analysts with basic SAS programming experience. The issues and techniques discussed in this course are directed toward database marketing, credit risk evaluation, fraud detection and other predictive modelling applications from banking, financial services, direct marketing, insurance, and telecommunications.

Objectives

This course covers predictive modelling using SAS/STAT^® with emphasis on the LOGISTIC procedure. It also discusses selecting variables, assessing models, treating missing values, and efficiency techniques for massive data sets.

After completing this course, you should be able to:

• fit logistic regression models for various sampling designs
• select relevant and non-redundant predictors
• impute missing values
• evaluate classifier performance
• score new cases

Prerequisite Skills

Before attending this course, you should:

• be able to execute SAS programs and create SAS data sets
• be familiar with probability theory and statistical modelling using SAS
• some familiarity with array processing and macro variables in SAS will be advantageous.

SAS^® System Modules used

SAS/STAT^®

Course Topics

Predictive Modelling

• Business applications
• Analytical challenges

Fitting the Logistic Regression Model

• Parameter estimation
• Adjustments for oversampling

Preparing the Input Variables

• Dealing with missing values
• The problems of categorical inputs
• Using variable clustering
• Using subset selection

Classifier Performance

• Looking at ROC curves and Lift charts
• Calculating optimal cutoffs
• Analysing K-S statistics

Evaluating Many Models

Non-linearities and Interactions

• Detection and Advanced Modelling Techniques

Reserve your place on this Newtyne SAS Training Course