The course looks at the theoretical and practical implications of a wide array of clustering techniques currently available in SAS. The techniques considered include cluster preprocessing, variable clustering, k-means clustering, and hierarchical clustering.

Learn how to

• prepare and explore data for a cluster analysis
• distinguish among many different clustering techniques, making informed choices about which to use
• evaluate the results of a cluster analysis
• determine the appropriate number of clusters to retain
• profile and describe clustered observations
• score observations into clusters.

Who should attend

Intermediate or senior level statisticians, data analysts, and data miners

Formats available	Duration
Live Web Classroom:	4 half-day session(s)

Before attending this course, you should

• be able to execute SAS programs and create SAS data sets. You can gain this experience by completing the SAS Programming 1: Essentials course.
• have completed a graduate-level course in statistics or the Statistics 1: Introduction to ANOVA, Regression, and Logistic Regression course.
• have an understanding of matrix algebra.

This course addresses SAS/STAT software.

Introduction to Clustering

• identifying types of clustering
• measuring similarity
• classification performance

Preparation for Clustering

• preparing data for cluster analysis
• using variable clustering for variable selection
• using graphical clustering aids
• making elongated clusters more spherical
• viewing the impact of input standardization

Partitive Clustering

• k-means clustering using PROC FASTCLUS
• outline the advantages of nonparametric clustering
• introducing PROC MODECLUS

Hierarchical Clustering

• comparing hierarchical clustering methods

Assessing Clustering Results

• determining the number of clusters in hierarchical and k-means clustering
• profiling a cluster solution
• scoring new observations

Cluster Analysis Case Study

• variable selection
• graphical exploration of selected variables
• hierarchical clustering and determining the number of clusters
• profiling the seven-cluster solution
• modeling cluster membership
• scoring the customer database

Canonical Discriminant Analysis (CDA)Plots

• using canonical discriminant analysis to summarize multivariate data
• interpret CANDISC procedure output

Fuzzy Clustering

• performing fuzzy clustering using the (Q-technique) FACTOR procedure
• interpreting PROC FACTOR output

Assessing Multivariate Normality

• assessing multivariate normality