Business Knowledge Series course

Presented by Donald Wedding, PhD, Consultant

This course focuses on best practices when doing cluster analysis, particularly when applied to CRM. However, the techniques presented are easily applicable to other analytic fields. Students with different backgrounds and skill levels will benefit from the course, which explores theory, practice, and application.

Learn how to

• Use clustering and segmentation.
• Determine the number of clusters present in data.
• Find the most effective clusters.
• Apply various approaches to clustering, including K-Means and Kohonen/SOM networks.
• Assess the quality of clusters.
• Use clustering in CRM, risk, fraud, and predictive models.
• Determine which variables should be used in clustering.
• Incorporate categorical and binary variables in clustering.
• Approximate clusters with decision trees.
• Incorporate soft “fuzzy” clustering.

Who should attend

Analysts from different backgrounds and with varying skill levels who work with large data sets

Formats available	Standard duration
Classroom:	2.0 days

Before attending this course, you should have some experience with the SAS programming language and SAS Enterprise Miner, as well as some knowledge about how to develop predictive models. Some experience with clustering is helpful but not required. Managers should have a basic understanding of analytics and how it is applied within their organization.

This course addresses Base SAS, SAS Enterprise Miner software.

Clustering Overview

• Introduction.
• Rules of clustering: Lesson 1 summary.

Cluster Types

• Introduction
• Rules of clustering: Lesson 2 summary.

Distance Metrics – Normalize/Standardize Data

• Distance metrics.
• Distance metric variables and scale.
• Transform to similar scales.
• Transform variable example.
• Rules of clustering: Lesson 3 summary.

Distance Metrics – Data Cleansing

• Outliers.
• Handling outliers.
• Class variables.
• Missing data.
• Rules of clustering: Lesson 4 summary.

Clustering Algorithms: Results

• Clustering results are inconsistent.
• Cluster seeds.
• Rules of clustering: Lesson 5 summary.

Different Clustering Algorithm Results

• SAS code used in examples.
• Random seeds example.
• Cluster seed selection methods.
• SAS Enterprise Miner seed selection methods.
• SAS Enterprise Miner – how many clusters?
• Rules of clustering: Lesson 6 summary.

Monte Carlo Cluster Selection

• Introduction.
• Save_Cluster_Info macro.
• Cluster node data collection example.
• Rules of clustering: Lesson 7 summary.

Assessing Clusters

• Size.
• Story.
• Sense.
• Stable.
• Satisfied.
• Rules of clustering: Lesson 8 summary.

Continuous Categorical Variables

• Introduction.
• Demographic clusters.
• Approximating clusters.
• Prospecting trees.
• Identifying a specific class.
• Rules of clustering: Lesson 9 summary.

Kohonen/SOM Clusters

• Introduction.
• Kohonen/SOM architecture.
• Kohonen/SOM example.
• Kohonen/SOM Monte Carlo clustering.
• Rules of clustering: Lesson 10 summary.
• Rules.

Fuzzy Clusters

• Boolean and fuzzy logic.
• Hard/Fuzzy clustering.
• Fuzzy cluster macro.
• Example of fuzzy cluster approximation.
• Application of fuzzy clusters.
• Rules of clustering: Lesson 11 summary.

Hierarchical Clusters

• Hierarchical clusters overview.
• Hierarchical cluster Example 1.
• How many clusters.
• SAS code for hierarchical clustering.
• Hierarchical cluster Example 2: Lesson 7 data set.
• Incorporating hierarchical clusters into SAS Enterprise Miner.
• Using hierarchical clusters in Monte Carlo clustering.
• Rules of clustering: Lesson 12 summary.